Introduction

Robot Statistics: The field of robotics has undergone remarkable advancements in recent years. Revolutionizing industries, shaping economies, and transforming the way we live and work.

Robots, once confined to the realms of science fiction, have become a tangible reality in our modern world.

These machines, capable of carrying out tasks autonomously or semi-autonomously, have found applications in diverse sectors. From manufacturing and healthcare to agriculture, transportation, and beyond.

Energy and utilities and industrial automation had the highest use of robotics in 2025, with over ** percent of enterprises using robots. Both are industries with planned and formulaic processes, plenty of tasks dangerous for humans, and easy integration of robotics into multiple tasks.

Introduction

Educational Robots Statistics: Educational robots are specialized devices employed in the educational field to engage students and facilitate learning. Especially in science, technology, engineering, and mathematics (STEM).

These robots possess the capability to be programmed, feature sensors, and are often mobile, allowing them to interact with their surroundings.

They are available in various forms, ranging from DIY robotic kits to pre-programmed and remotely controlled robots, serving as hands-on learning aids.

Educational robots find widespread use in STEM education, coding instruction, and problem-solving tasks. Delivering practical knowledge and preparing students for future careers in technology-related professions.

While they offer advantages such as improved learning and the development of critical skills. Challenges like cost, teacher training, and maintenance should be considered.

China had by far the highest use of robotics among the countries surveyed, with ** percent of enterprises using robotics in 2025. In all other surveyed countries, robotics are used by between - percent of enterprises. The reasons behind this are numerous but include China's demographic issues, producing huge quantities of robots itself, as well as being a base manufacturing powerhouse - a sector where robotics usage is growing fastest and easiest to implement.

Conventionally, the kinematic structure of a robot is assumed to be known and data from external measuring devices are used mainly for calibration. We take an agent-centric perspective to explore whether a robot could learn its body structure by relying on scarce knowledge and depending only on unorganized proprioceptive signals. To achieve this, we analyze a mutual-information-based representation of the relationships between the proprioceptive signals, which we call proprioceptive information graphs (pi-graph), and use it to look for connections that reflect the underlying mechanical topology of the robot. We then use the inferred topology to guide the search for the morphology of the robot; i.e. the location and orientation of its joints. Results from different robots show that the correct topology and morphology can be effectively inferred from their pi-graph, regardless of the number of links and body configuration., The datasets contain the proprioceptive signals for a robot arm, a hexapod, and a humanoid, including joint position, velocity, torque, body angular and linear velocities, and body angular and linear accelerations. The robot manipulator experiment used simulated robot joint trajectories to generate the proprioceptive signals. These signals were computed using the robot's Denavit-Hartenberg parameters and the Newton-Euler method with artificially added noise. In the physical experiment, joint trajectories were optimized for joint velocity signal entropy, and measurements were obtained directly from encoders, torque sensors, and inertial measurement units (IMU). In the hexapod and humanoid robot experiments, sensor data was collected from a physics simulator (Gazebo 11) using virtual IMU sensors. Filters were applied to handle measurement noise, including low-pass filters for offline estimation and moving average filters for online estimation, emphasizing noise reduction for angular veloc..., , # Machine Learning Driven Self-Discovery of the Robot Body Morphology

The repository contains:

• Data sets
• Links to MATLAB source code

Requirements

• MATLAB's Robotics System Toolbox
• MATLAB's Optimization Toolbox
• ToolÂboxes for optiÂmization on manifolds and matrices MANOPT
• Java Information Dynamics Toolkit JIDT

NOTE: MATLAB 2021b was used.

Sharing/Access information

All datasets are also publicly available at Kaggle; these are the corresponding links:

• Simulated robot manipulator with fixed and moving base here
• Physical manipulator experiment (fixed base) here
• Simulated hexapod robot here
• Simulated humanoid robot [h...

Simulation plays a crucial role in modern academic study, particularly in the field of artificial intelligence (AI). The simulation environment can mimic real-world scenarios, allowing the AI agent to learn, adapt, and make decisions in a controlled and safe setting. This thesis tackles two important problems in building the next generation of artificial general intelligence (AGI): how to efficiently train an AI agent with values and how to overcome the simulation to reality gap to bring the training results to real-world applications. The current studies of AI mainly consider learning about the potential or energy function (U), referring to understanding the impact of the outside environment. The U function helps the agent apprehend the physical world laws, natural potentials, and social norms. However, taking into account the value learning, usually representing modeling one's inside thinking, benefits the agent to derive its goals, intents, and social values. Our research shows that both U and V learning are equally important to the pathway to AGI. The learning of U is usually data-driven. It enables the agent to imitate and complete the task through statistical learning. By incorporating the value function, the agent can spontaneously specify a task plan and its behavior is more in line with human cognition and value.This thesis consists of three parts: (1) Potential function learning, which explores the process of acquiring knowledge or skills that are useful and practical for a particular purpose. (2) Value learning when learning the potential (U) function can not satisfy all the learning goals, which investigates situations where utility-based learning approaches might be limited or ineffective. (3) Combining U and V learning, which focuses on the integration of simulation-based learning and data-driven learning methods. We primarily focus on assessing the effectiveness of U learning within a simulated environment. Our investigation commences with agents operating in a controlled simulated setting, where the action space is intentionally kept small. Through rigorous testing and iterative refinement, we gradually expand the scope of our analysis to encompass agents dealing with increasingly complex and continuous action spaces. Upon achieving compelling results in the simulated realm, we proceed to the crucial next step: transferring the knowledge and expertise gained from the well-trained agents in the simulation space to real-world scenarios. This process entails adapting the learned policies, strategies, and decision-making capabilities of the agents to navigate the intricacies and uncertainties of genuine environments.

This is an example dataset recorded using version 1.0 of the open-source-hardware OpenAXES IMU. Please see the github repository for more information on the hardware and firmware. Please find the most up-to-date version of this document in the repository

This dataset was recorded using four OpenAXES IMUs mounted on the segments of a robot arm (UR5 by Universal Robots). The robot arm was programmed to perform a calibration movement, then trace a 2D circle or triangle in the air with its tool center point (TCP), and return to its starting position, at four different speeds from 100 mm/s to 250 mm/s. This results in a total of 8 different scenarios (2 shapes times 4 speeds). The ground truth joint angle and TCP position values were obtained from the robot controller. The calibration movement at the beginning of the measurement allows for calculating the exact orientation of the sensors on the robot arm.

The IMUs were configured to send the raw data from the three gyroscope axes and the six accelerometer axes to a PC via BLE with 16 bit resolution per axis and 100 Hz sample rate. Since no data packets were lost during this process, this dataset allows comparing and tuning different sensor fusion algorithms on the recorded raw data while using the ground truth robot data as a reference.

In order to visualize the results, the quaternion sequences from the IMUs were applied to the individual segments of a 3D model of the robot arm. The end of this kinematic chain represents the TCP of the virtual model, which should ideally move along the same trajectory as the ground truth, barring the accuracy of the IMUs. Since the raw sensor data of these measurements is available, the calibration coefficients can also be applied ex-post.

Since there are are 6 joints but only 4 IMUS, some redundancy must be exploited. The redundancy comes from the fact that each IMU has 3 rotational degrees of fredom, but each joint has only one:

• The data for q0 and q1 are both derived from the orientation of the "humerus" IMU.
• q2 is the difference^† between the orientation of the "humerus" and "radius" IMUs.
• q3 is the difference between the orientation of the "radius" and "carpus" IMUs.
• q4 is the difference between the orientation of the "carpus" and "digitus" IMUs.
• The joint q5 does not influence the position of the TCP, only its orientation, so it is ignored in the evaluation.
• ^† Of course, difference here means not the subtraction of the quaternions but the rotational difference, which is the R1 * inv(R0) for two quaternions (or rotations) R0 and R1. The actual code works a bit differently, but this describes the general principle.

Data

• Data recorded from the IMUs is in the directory measure_raw-2022-09-15/, one folder per scenario. In those folders, there is one CSV file per IMU.
• Data recorded from the robot arm is in the directory measure_raw-2022-09-15/robot/, one CSV and MAT file per scenario.
• Some photos and videos of the recording process can be found in Media. Videos are stored in git lfs.

Evaluation

The file openaxes-example-robot-dataset.ipynb is provided to play around with the data in the dataset and demonstrate how the files are read and interpreted. To use the notebook, set up a Python 3 virtual environment and therein install the necessary packets with pip install -r resuirements.txt. In order to view the graphs contained in the ipynb file, you will most likely have to trust the notebook beforehand, using the following command:

jupyter trust openaxes-example-robot-dataset.ipynb

Beware: This notebook is not a comprehensive evaluation and any results and plots shown in the file are not necessarily scientifically sound evidence of anything.

The notebook will store intermediate files in the measure_raw-2022-09-15 directory, like the quaternion files calculated by the different filters, or the files containing the reconstructed TCP positions. All intermediate files should be ignored by the file measure_raw-2022-09-15/.gitignore.

The generated intermediate files are also provided in the file measure_raw-2022-09-15.tar.bz2, in case you want to inspect the generated files without running the the notebook.

Tools

A number of tools are used in the evaluation notebook. Below is a short overview, but not a complete specification. If you need to understand the input and output formats for each tool, please read the code.

• The file calculate-quaternions.py is used in the evaluation notebook to compute different attitude estimation filters like Madgwick or VQF on the raw accelerometer and gyroscrope measurements at 100 Hz.
• The directory madgwick-filter contains a small C program that applies the original Madgwick filter to a CSV file containing raw measurements and prints the results. It is used by calculate-quaternions.py.
• The file calculate-robot-quaternions.py calculates a CSV file of quaternions equivalent to the IMU quaternions from a CSV file containing the joint angles of the robot.
• The program dsense_vis mentioned in the notebook is used to calculate the 3D model of the robot arm from quaternions and determine the mounting orientations of the IMUs on the robot arm. This program will be released at a future date. In the meantime, the output files of dsense_vis are provided in the file measure_raw-2022-09-15.tar.bz2, which contains the complete content of the measure_raw-2022-09-15 directory after executing the whole notebook. Just unpack this archive and merge its contents with the measure_raw-2022-09-15 directory. This allows you to explore the reconstructed TCP files for the filters implemented at the time of publication.

Introduction

Robotics Industry Statistics: The robotics industry has rapidly transformed from a futuristic vision into a core part of today’s industrial operations. As of 2024, around 3.4 million industrial robots are in use worldwide, performing tasks in manufacturing, logistics, healthcare, and even domestic environments. In automotive factories, robots handle nearly 50% of production processes. The global robotics market, including both industrial and service robots, is projected to exceed USD 45 billion in 2025, driven by increased automation demand.

In 2023, more than 550,000 new robots were installed globally, setting a record for annual deployment. The adoption of collaborative robots (cobots) also grew by over 20% year-over-year. Robots are now not only assembling vehicles but also assisting in surgical procedures, warehouse management, and household chores. In Japan alone, over 350,000 industrial robots are operational, reflecting the country’s leadership in automation. Meanwhile, China accounts for nearly 52% of all global robot installations, highlighting its rapid industrial scaling.

This paper presents updated statistics and trends from 2024 and 2025, providing a numerical overview of robotics integration across industries. So let's delve into some interesting statistics to get a better sense of the size and growth of the robotics industry.

Survey information collected for the study entitled "Expectations and Perceptions of Healthcare Professionals for Robot Deployment in Hospital Environments during the COVID-19 Pandemic"This dataset is shared under the following license: Creative Commons — Attribution-NonCommercial-ShareAlike 4.0 International — CC BY-NC-SA 4.0

This statistic displays the global spending on robotics from 2000 to 2025. In 2015, spending on industrial robotics came to about 11 billion U.S. dollars. Industrial robots are increasingly used in the automotive sector, as well as the metal manufacturing and materials handling industries.

This element involves the development of software that enables easier commanding of a wide range of NASA relevant robots through the Robot Application Programming Interface Delegate (RAPID) robot messaging system and infusing the developed software into flight projects.  In June and July of 2013, RAPID was tested on ISS as the robot messaging software for the Technology Demonstration Mission (TDM) Human Exploration Telerobotics (HET) Surface Telerobotics experiment.  RAPID has also been made available to — and integrated with — the Robot Operating System (ROS), a popular software framework for developing state-of-the-art robots for ground and space. While ROS powers a number of new robots and components such as Robonaut 2’s climbing legs and R5, the addition of RAPID allows these robots to interoperate in collaborative human-robot teams, safely and effectively over time-delayed communications links. The objective this year is to take this space-tested software and extend it to providing video streaming from remote robots and delivering this new capability to the Exploration Ground Data Systems (xGDS) area within HRS.  xGDS will then deliver its software to Science Mission Directorate (SMD) funded field tests to improve the technology readiness moving leading (potentially) to being used for the Lunar Prospector Mission ground data systems.  Success will involve delivering RAPID to xGDS and then xGDS supporting SMD field test.

The team is also developing algorithms for sensors capable of reconstructing remote worlds and efficiently shipping that remote environment back to earth using the RAPID robot messaging system.  This type of system could eventually lead to scientists on earth gain new insights as they are able to step into the remote world.  This sensor also has the ability to engage the public, bringing remote worlds back to earth.  During FY13, this task used science operations personnel from current SMD projects to objectively measure improvement in remote science target selection and decision-making based. The team continues to work with SMD projects to ensure that the technologies being developed are directly responsive to SMD project personnel needs. The objective of this work in FY14 is to expand the range of science operations tasks addressed by the technology, and to perform laboratory demonstrations for JPL/SMD stakeholders of the immersive visualization of data from a sensor using an SMD representative environment.

During 2014, the “Controlling Robots Over Time Delay” project element will develop two technologies:

• Develop RAPID robot messaging for unified cross-center operations platform for TDM, xGDS, and CCSDS
• Sensor Systems for the Construction of Immersive Virtual Environments

Introduction

Agricultural Robots Statistics: The transition that agriculture, the backbone of human civilization, is currently experiencing is largely driven by advanced technology. Among the technological innovations in agriculture, agricultural robots—often referred to as "bots"—are at the forefront of this change, disrupting traditional farming practices. Agricultural robots are autonomous machines that perform various tasks, including planting, harvesting, monitoring crop health, and managing livestock.

Their increasing acceptance is due to their ability to enhance efficiency, address the shortage of agricultural labor, and meet the rising demand for food with higher productivity. This article will present statistics on agricultural robots for 2025, highlighting key trends in the market and the factors contributing to their growth.

The dataset contains both the robot's high-level tool center position (TCP) health data and controller-level components' information (i.e., joint positions, velocities, currents, temperatures, currents). The datasets can be used by users (e.g., software developers, data scientists) who work on robot health management (including accuracy) but have limited or no access to robots that can capture real data. The datasets can support the: Development of robot health monitoring algorithms and tools Research of technologies and tools to support robot monitoring, diagnostics, prognostics, and health management (collectively called PHM) Validation and verification of the industrial PHM implementation. For example, the verification of a robot's TCP accuracy after the work cell has been reconfigured, or whenever a manufacturer wants to determine if the robot arm has experienced a degradation. For data collection, a trajectory is programmed for the Universal Robot (UR5) approaching and stopping at randomly-selected locations in its workspace. The robot moves along this preprogrammed trajectory during different conditions of temperature, payload, and speed. The TCP (x,y,z) of the robot are measured by a 7-D measurement system developed at NIST. Differences are calculated between the measured positions from the 7-D measurement system and the nominal positions calculated by the nominal robot kinematic parameters. The results are recorded within the dataset. Controller level sensing data are also collected from each joint (direct output from the controller of the UR5), to understand the influences of position degradation from temperature, payload, and speed. Controller-level data can be used for the root cause analysis of the robot performance degradation, by providing joint positions, velocities, currents, accelerations, torques, and temperatures. For example, the cold-start temperatures of the six joints were approximately 25 degrees Celsius. After two hours of operation, the joint temperatures increased to approximately 35 degrees Celsius. Control variables are listed in the header file in the data set (UR5TestResult_header.xlsx). If you'd like to comment on this data and/or offer recommendations on future datasets, please email guixiu.qiao@nist.gov.

Offline reinforcement learning (RL) is a promising direction that allows RL agents to be pre-trained from large datasets avoiding recurrence of expensive data collection. To advance the field, it is crucial to generate large-scale datasets. Compositional RL is particularly appealing for generating such large datasets, since 1) it permits creating many tasks from few components, and 2) the task structure may enable trained agents to solve new tasks by combining relevant learned components. This submission provides four offline RL datasets for simulated robotic manipulation created using the 256 tasks from CompoSuite Mendez et al., 2022. In every task in CompoSuite, a robot arm is used to manipulate an object to achieve an objective all while trying to avoid an obstacle. There are for components for each of these four axes that can be combined arbitrarily leading to a total of 256 tasks. The component choices are * Robot: IIWA, Jaco, Kinova3, Panda* Object: Hollow box, box, dumbbell, plate* Objective: Push, pick and place, put in shelf, put in trashcan* Obstacle: None, wall between robot and object, wall between goal and object, door between goal and object The four included datasets are collected using separate agents each trained to a different degree of performance, and each dataset consists of 256 million transitions. The degrees of performance are expert data, medium data, warmstart data and replay data: * Expert dataset: Transitions from an expert agent that was trained to achieve 90% success on every task.* Medium dataset: Transitions from a medium agent that was trained to achieve 30% success on every task.* Warmstart dataset: Transitions from a Soft-actor critic agent trained for a fixed duration of one million steps.* Medium-replay-subsampled dataset: Transitions that were stored during the training of a medium agent up to 30% success. These datasets are intended for the combined study of compositional generalization and offline reinforcement learning. Methods The datasets were collected by using several deep reinforcement learning agents trained to the various degrees of performance described above on the CompoSuite benchmark (https://github.com/Lifelong-ML/CompoSuite) which builds on top of robosuite (https://github.com/ARISE-Initiative/robosuite) and uses the MuJoCo simulator (https://github.com/deepmind/mujoco). During reinforcement learning training, we stored the data that was collected by each agent in a separate buffer for post-processing. Then, after training, to collect the expert and medium dataset, we run the trained agents for 2000 trajectories of length 500 online in the CompoSuite benchmark and store the trajectories. These add up to a total of 1 million state-transitions tuples per dataset, totalling a full 256 million datapoints per dataset. The warmstart and medium-replay-subsampled dataset contain trajectories from the stored training buffer of the SAC agent trained for a fixed duration and the medium agent respectively. For medium-replay-subsampled data, we uniformly sample trajectories from the training buffer until we reach more than 1 million transitions. Since some of the tasks have termination conditions, some of these trajectories are trunctated and not of length 500. This sometimes results in a number of sampled transitions larger than 1 million. Therefore, after sub-sampling, we artificially truncate the last trajectory and place a timeout at the final position. This can in some rare cases lead to one incorrect trajectory if the datasets are used for finite horizon experimentation. However, this truncation is required to ensure consistent dataset sizes, easy data readability and compatibility with other standard code implementations. The four datasets are split into four tar.gz folders each yielding a total of 12 compressed folders. Every sub-folder contains all the tasks for one of the four robot arms for that dataset. In other words, every tar.gz folder contains a total of 64 tasks using the same robot arm and four tar.gz files form a full dataset. This is done to enable people to only download a part of the dataset in case they do not need all 256 tasks. For every task, the data is separately stored in an hdf5 file allowing for the usage of arbitrary task combinations and mixing of data qualities across the four datasets. Every task is contained in a folder that is named after the CompoSuite elements it uses. In other words, every task is represented as a folder named

The Robot Motion Dataset contains experimental data from a study investigating human-robot interaction with a leader who used a teleoperated follower robot. So, thirty participants controlled the TFR wearing a virtual reality (VR) headset and using rudder platform. Thus, the dataset includes sensor data such as accelerometer, gyroscope, trajectory, objects' distance, data questionnaires, and so forth.

The dataset was used in the work presented in the following article: ASAP.

Introduction

Robotic Process Automation Statistics: RPA is a transformative technology that leverages robot software to automate rule-based tasks within digital systems. It operates by identifying repetitive tasks and developing software bots to execute them.

Seamlessly integrating these bots with existing software applications. RPA offers numerous benefits, including cost efficiency, accuracy, scalability, and enhanced productivity.

Its adoption is on the rise across industries, with the global RPA market poised for significant growth. This technology has the potential to revolutionize business operations.

By reducing costs, improving efficiency, and allowing human employees to focus on more strategic activities. Ultimately enhancing overall productivity and competitiveness.

This is a collection of data used for the article "Indicators for the use of Robotic Labs in Basic Biomedical Research". The primary result files are: -metamapMethods.csv -sodaMethods.csv

These were generated by running descriptive statistics of annotating the papers listed in articles_piis_dois.csv with the Medical Subject Headings 2015 vocabulary.

Stage tasks: Task 1: Development of algorithms for statistical analysis of attribute values for data purification. The aim of the task was to develop an algorithm that is able to identify the type of attribute (scalar, discrete) and depending on the type (text, number, date, text label, etc.) and deduce which values can be considered correct and which are incorrect and cause noise dataset, which in turn affects the quality of the ML model. Task 2: Development of algorithms for statistical analysis of data attributes in terms of optimal coding of learning vectors. The aim of the task was to develop an algorithm that is able to propose optimal coding of the learning vector to be used in the ML process and perform the appropriate conversion, depending on the type (text, number, date, text label, etc.) for each type of attribute (scalar, discrete). e.g. converting text to word instance matrix format. It was necessary to predict several possible conversion scenarios that are most often used in practice, resulting from the heuristic knowledge of experts. Task 3: Developing a prototype of an automatic data cleaning and coding environment and testing the solution on samples of production data. Industrial Research: Task No. 2. Research on the meta-learning algorithm Task 1: Review of existing meta-learning concepts and selection of algorithms for further development The aim of the task was to analyze the state of knowledge on meta-learning in terms of the possibility of using existing research results in the project - a task carried out in the form of subcontracting by a scientific unit. Task 2: Review and development of the most commonly used ML algorithms in terms of their susceptibility to hyperparameter meta-learning and practical usefulness of the obtained models. The aim of the task was to develop a pool of basic algorithms that will be used as production algorithms, i.e. performing the right predictions. The hyperparameters of these algorithms have been meta-learning. It was therefore necessary to develop a model of interaction of the main algorithm with individual production algorithms. – task carried out in the form of subcontracting by a scientific unit. Task 3: Development of a meta-learning algorithm for selected types of ML models The aim of the task was to develop the main algorithm implementing the function of optimizing hyperparameters of production models. It should be noted that the hyperparameters have a different structure depending on the specific production model, so the de facto appropriate solution was to use a different optimization algorithm for each model separately. Task 4: Developing a prototype of the algorithm and testing the operation of the obtained production data models. Experimental development work: Task No. 3. Research on the prototype of the architecture of the platform implementation environment Task 1: Developing the architecture of the data acquisition and storage module. The aim of the task was to develop an architecture for a scalable ETL (Extract Transform Load) solution for efficient implementation of the source data acquisition process (Data Ingest). An attempt was made to consider appropriate parsing algorithms and standardization of encoding data of various types (e.g. dates, numbers) in terms of effective further processing. Task 2: Development of a module for configuring and executing data processing pipelines in a distributed architecture. Due to the high complexity of the implemented algorithms, it was necessary to develop an architecture that would allow pipeline processing of subsequent data processing steps on various machines with the possibility of using a distributed architecture in a cloud and/or virtual environment. The use of existing concepts of distributed architectures, such as Map Reduce, was considered here. Task 3: Development of a user interface enabling intuitive control of data processing.

The sales value of new robot installations in the global food and beverages industry is expected to drop to some 426 million U.S. dollars in 2020. However, the sales value should bounce back in the following years, peaking at around 523 million U.S. dollars in 2022. The demand for industrial robots in the industry is forecast to increase in China, Japan, and the United States, while in the rest of the world the demand will either be stagnating or decreasing over the forecast period.

The intend of this data set is the cooperation within SimTech. It will be particularly interesting for data-based modeling and control which is a key area of the research of project network 4. We are proud to provide real-world data, which is essential for the benchmark of any data-based method. Additionally, we are able to provide reference solutions in order to evaluate the predictive quality of the methods tested. Finally, an example on how this data set can be used with Gaussian process (GP) regression in order to predict the systematic mismatches of the mobile robot is given. The data set contains input-output data of an omnidirectional mobile robot. The inputs to the mobile robots are the desired speeds in the plane as well as an angular velocity of the robot around its vertical axis. The corresponding outputs are the position in the plane and the robots orientation in an inertial frame of reference. The data set is provided in the Matlab *.mat format.