As robotics continues to expand to more domains, the development and maintenance of suitable experimental facilities are becoming bottlenecks in the innovation process.
In fact, there is a significant gap today between the theoretical foundations that are being broadly pursued and a focused, application-driven transition from small-scale experiments to robust and high-impact deployments. This gap is both scientific and practical.
The solution is to allow researchers from different institutions, disciplines, and backgrounds to come together around a standard testbed consisting of several labs serving as a resource for research in robotics for advanced manufacturing and material handling. This can accelerate innovation and build on past findings more effectively than what is currently done.
The testbed contains representative state-of-the-art manufacturing robots, advanced multi-fingered grippers, sensors, conveyors, and an industrial robot arm mounted on a linear rail or on a pedestal. A custom-configured automatic guided vehicle (AGV) is used to research industrial, vehicular safety, and performance standards, including mobile manipulation. The robot systems also include vision and force-torque sensing capability. Highly accurate simulation systems model the essential components and allow different combinations of real and virtual features to be used in experiments.
However, the development and maintenance of meaningful, large-scale robotic testbeds is a resource-intensive undertaking, which is why it is particularly well-suited to a shared and even remote-access format.
Some specialized robot testbeds exist (such as UMass Lowell’s NERVE Center, the Southwest Research Institute’s Small Robotic Vehicle Evaluation and Applications Group, response robots, and manufacturing robots at NIST in Gaithersburg, Texas A&M’s Disaster City). Still, these testbeds lack a variety of robot systems, causing the researchers to bring their own robot systems, limiting the ability to test the generalizability of algorithms and preventing people without robot systems from testing their theoretical results in the real world.
To accelerate the development and practical testing of robot systems, shared community resources of testbeds for various application domains with multiple robot systems must be developed, each with a particular application focus (e.g., agriculture, marine, manufacturing, medical).
To maximize the use of available resources, existing facilities can be expanded to create a comprehensive shared infrastructure (i.e., one with both testbeds and shared robot systems) while developing testbeds for application domains where no such facilities yet exist.
1. Flexible Research Platforms
It’s critical for a remote-access research testbed to be structured in a way that allows multiple research questions and experiments to be pursued to be truly useful. Furthermore, the testbed must evolve over time to stay current with changing research trends and directions. It must also be possible to automatically specify experimental setups and scenarios, which necessitates research into modular, interoperable hardware (plug-and-play, for example) and software (robotic experiment description languages, for example) to allow for their integration into larger eco-systems and downstream commercialization.
Furthermore, because infrastructure is a community resource, each facility should have a user committee to allocate site usage and make suggestions for facility improvements. Finally, while each facility will focus on a different application, there should be a collaboration between them to avoid duplication of effort and share best practices.
2. Community Consensus Validation Benchmark Frameworks
Several research groups have developed proprietary methods for evaluating the quantitative performance of robotic systems and the human-robot interaction. Individual groups have started gathering and sharing their data sets and best practices. However, due to a lack of realistic and relevant test environments, efforts are fragmented and disconnected (physical and virtual benchmarks). It’s critical to have a multifaceted validation strategy (for example, supporting virtual and physical testing; staged evaluation of components, subsystems, and systems; device vs. user testing).
The development of open-access frameworks for creating, collecting, and curating appropriate reference environments and data corpora against which quantitative performance can be assessed would significantly speed up the technology development and transfer process. Previous efforts in the robotics community have bolstered the case that posing these as a competition or grand challenge could help focus the energy of both the academic and industrial communities while also paving the way for future standardization efforts. Through ASTM, IEEE, and ASME, some robotics domains have already developed standard test methods and metrics; these efforts should continue as new application domains are developed. The Robotics-VO could oversee community discussions aimed at developing these shared resources.
3. Reference Open-Access Testbeds
The exponential growth of robotics has resulted in an explosion in the number and variety of solutions available. For example, the variety of commercially available manipulator arms, mobile bases, and grippers can be overwhelming. It is becoming increasingly difficult for a researcher to grasp algorithms to gain access to various manipulators with various grippers to evaluate the effectiveness of their algorithms. The lack of truly industry-grade testbeds with interoperable hardware and software modules is starting to stifle innovation.
Efforts to develop open-source platforms are a good place to start. For example, the ROS Framework is an excellent example of robotics software. The availability of such plug-and-play frameworks will allow research groups to concentrate on their specific subtopics while still contributing to a larger, more cohesive community effort. Furthermore, system interoperability and synergistic technical tools (e.g., programming, hardware, communication) are critical for accelerating robotic system research and development and will benefit academia and industry.
