ARM Institute: 25-01 Funding Project Calls

Information Webinar: March 13, 2025 (https://arminstitute.zoomgov.com/webinar/register/WN_D5Osha6DQe-mfYaRBB7epQ#/registration)

Concept Paper Submission Deadline: April 9, 2025

$3 Million in funding available!

Must be a member organization to apply

Membership Information here: https://arminstitute.org/membership/

Application Info here: Project Calls - ARM Institute

The ARM Institute has released its 25-01 Technology Project Call.

Special Topic Areas include:

• Special Topic Area 1 (STA1): Multi-modal inputs for AI robotics in manufacturing
• Special Topic Area 2 (STA2): ?Rapid Re-Tasking and Robot Agility
• Special Topic Area 3 (STA3): ?Multi-Robot, Multi-Human Collaboration
• Special Topic Area 4 (STA4): Adaptive Real-Time Path Planning and Control

Program Description

The ARM Institute is soliciting submissions that respond to specific needs of the manufacturing industry and the Department of Defense (DoD). This project call is seeking projects to make focused investments in the Special Topic Areas (STA) described below. These Special Topic Areas have been derived from the institute’s prior research, with guidance from the Institute’s Technology Advisory Committee (TAC) and partners. Project teams are encouraged to leverage successfully completed ARM Institute technology development programs as well as Consortium Developed Intellectual Property (CDIP) in their submissions.

Program Special Topic Areas

STA 1 - Multi-modal inputs for AI robotics in Manufacturing

Robots are gaining exceptional perception capabilities through recent advancements in Artificial Intelligence (AI). While deep learning methods in robotics have predominantly relied on RGB image data in the past, there is no reason to restrict their use only to vision data. Multimodal sensing involves the use of various sensors that capture diverse data types such as visual, auditory, tactile, and proprioceptive, to help robots perceive and interpret their surroundings.

Implementing multimodal sensing in robotics can significantly enhance their ability to make decisions and control various tasks. In manufacturing applications, where a vast range of sensor data such as force, pressure, temperature, hyperspectral imaging, and acoustic are used, multimodal sensing can prove especially beneficial. Manufacturing automation applications are increasingly recognizing the importance of multimodal sensing in improving the precision, efficiency, and safety of different processes.

Some examples of how multimodal sensing-based AI can be used in manufacturing robotics applications are listed below:

·???????? Object Recognition: To gather information about the objects in their surroundings, robots can utilize cameras, 3D scanners, and various other sensors. In addition, haptic feedback and data regarding an object's texture, shape, and hardness can be obtained using tactile sensors.

·???????? Navigation and Mapping: Sensors like LIDAR, sonar, and radar enable robots to produce 3D maps of their surroundings and identify any obstacles. Moreover, robots can estimate their location and orientation by utilizing GPS, IMU, and wheel encoders.

·???????? Human-Robot Interaction: By using sensors like cameras and microphones, robots can perceive and decipher human speech, facial expressions, and gestures. This ability empowers robots to comprehend and respond to human commands, engage with people more organically, and offer support in diverse fields.

·???????? Manipulation and Assembly: To perform safe manipulation and grasping operations, robots can use a combination of force/torque sensors, tactile sensors, and vision systems. These capabilities can encompass manipulating and handling objects of varying weights, shapes, and sizes. Assembly of parts can also be automated with the use of robots equipped with force/torque sensors and vision systems. These robots can detect and align parts with precision and apply the necessary force to guarantee proper fit and function.

·???????? Quality Control in Processing Applications: Maintaining specific quality standards is crucial in manufacturing. Multimodal sensing can aid in product inspection by employing diverse sensors like 3D scanners, cameras, and ultrasonic sensors. This approach can effectively identify defects like impurities, deformations, and cracks.

·???????? Safety: To enhance safety in automated manufacturing settings, multimodal sensing can be utilized to identify potential hazards such as collisions, extreme vibrations, and temperature fluctuations. A combination of sensors including gas sensors, infrared thermometers, and vibration sensors can be employed to observe the environment and halt the processing work cell in the event of any irregularities.

Beyond RGB images, deep learning techniques can be highly beneficial for other forms of data as well. To achieve this, sophisticated perception technologies with a multi-modal approach are required. Currently, research is underway in the field of robotic pick-and-place applications to use multi-modal sensing. Since various sensors generate data at different frequencies, employing multiscale temporal methods will be necessary to manage this data effectively. Interpreting and understanding the results of multi-modal learning will be challenging since the models may be complex and difficult to explain. This area will require development of new methods. Combining multimodal data can be very challenging when data is incomplete or missing. We will need to develop methods that are robust to such problems.

STA 2 - Rapid Re-Tasking and Robot Agility

Advances in automation have provided for sustained productivity increases and manufacturing growth over the past decade. Sustaining this growth will require automation to become more agile and flexible, enabling the automation of tasks that require a high degree of human dexterity and the ability to react to unforeseen circumstances. Robot’s traditional program-by-teaching model takes considerable time, requires extensive expertise, and does not lend itself to tasks that require adaptability. This has limited robots to high-volume, repetitive operations and precluded them from low-volume, time critical, and flexible projects. Offline programming of robots is possible, like the computer-aided manufacturing (CAM) method widely used for machine tools. However, the poor accuracy of robots compared with machine tools limits them to jobs with low tolerance requirements, or requires additional methods such as calibration, modeling, and external sensing to improve their accuracy. These methods increase the upfront cost of a robotic system. However, advances and cost reduction in sensing technologies (especially laser scanning) have brought robot systems into the price range of even small-to-medium enterprises.

In this context, agility is defined broadly to address:

·???????? Failure identification and recovery, where robots can detect failures in a manufacturing process and automatically recover from those failures.

·???????? Automated planning, to minimize (or eliminate) the up-front robot programming time when a new product is introduced.

·???????? Fixtureless environment, where robots can sense the environment and perform tasks on parts that are not in predefined locations.

·???????? Plug and play robots, where robots from different manufacturers can be swapped in and out without the need for reprogramming.

Manufacturers have difficulty addressing this because of the complexity of robot systems, the lack of understanding of robot capabilities, and the absence of measurement science and tools to assess and assure the robot’s agility. Manufacturers have often avoided and/or accepted the agility limitations since robots are primarily used in large companies for large lot size production, and changeover is performed infrequently and manually. However, addressing the robot agility challenges will allow manufacturers to gain more value from their robots by more rapidly reconfiguring and re-tasking robot systems, make robots more accessible to small and medium organizations, and provide large organizations with greater efficiency in their operations.

Artificial Intelligence (AI) approaches can play a significant role. Traditional AI approaches applied to dynamic re-planning and reasoning over formal knowledge representations are being expanded into machine learning techniques applied to perception and path planning, with the goal of designing experiments that will yield AI training data.

Focus areas could include, but not be limited to, learning from teach tables, optimizing action sequences for shorter run times, and making the robots easier to configure/program for novices.

STA 3 - Multi-Robot, Multi-Human Collaboration

Future manufacturing environments will have multiple robots and humans working in close- proximity and collaborating on specific tasks. Robots are currently unable to reliably detect human presence in environments with high uncertainty and contextually understand human activity.

Similarly, no solutions currently exist for dynamic, distributed sensing for safety. ARM has identified the development of real-time coordination between multiple robots and multiple humans to perform advanced manufacturing processes as a key area of interest. A successful project may address one of the following:

·???????? Collaborative frameworks to process and assign tasks based on skills, capabilities, etc. Reduce human supervision/intervention for decision making and resolving conflicts, by increasing the accuracy of inference on the decision making and action planning. Reduce the required manual human orchestration (i.e., manual preprogramming) of reasoning (i.e., hard coded rules) on multi-robot tasks.

·???????? Technology to sense and interpret the actions and intentions of a human or robot, or to communicate planned robotic actions. Robots need to have safety-rated abilities to detect, localize, track, and work with humans. In instances where non-safety-rated sensors are used in safety-related applications, there is no mechanism by which the safety performance of these sensors can be assessed. This project call looks for solutions that advance the ability for robots to not only reliably and robustly detect dynamic objects in their environment but also differentiate humans and the actions they are performing with defined certainty.

·???????? Develop methods to bridge the gaps when transferring robot skills from simulation to reality, due to missing, new, or inaccurate data. Develop methods to infer explainable action plans for the robots to ensure safety and minimize disruption in the process.

·???????? Build robust, distributed world models of manufacturing environments with uncertainty derived from multiple mobile and static sensing agents. Fuse multiple sensor modalities attached to mobile agents within a single world model. Return that world model or decisions made from it to the individual agents to adapt their sensing activities in a dynamic manner. Demonstrate mapping of a “ground truth world”, which requires the multiple sensor modes to fully define.

·???????? Overcome challenges of robot/sensor localization with uncertainty, robust wireless communication technologies between heterogenous agents/sensors, Protocols for presenting sensor data of differing dimensionality, Quantification of uncertainty of sensor data, Validity/Interpolate-ability of data in world model (needs meta data from sensors)

STA 4 - Adaptive Real-Time Path Planning and Control

1.1.1.????? Adaptive real-time path planning and control is increasingly essential in manufacturing environments, particularly as industries adopt advanced automation and robotics. This technology optimizes the movement of machines, robots, and material handling systems, enhancing efficiency, productivity, and safety on the factory floor. A successful solution conveys a technical approach or methodology that enables advanced manufacturing robotic systems to adjust a path or trajectory to variations that may occur in the system or process. Projects shall include the following in effort to develop adaptive real-time path planning and control capability:

·???????? Dynamic Environment Adaptation: Systems can continuously assess and respond to changes in the manufacturing environment, such as new obstacles, changes in production schedules, or equipment availability.

·???????? Adaptive Algorithms: Advanced algorithms enable real-time adjustments to paths based on current conditions, optimizing routes for speed and minimizing delays while ensuring safety in high-traffic areas.

·???????? Multi-Robot Coordination: Adaptive path planning facilitates effective collaboration among multiple robots or automated guided vehicles (AGVs), improving throughput and reducing the risk of collisions.

Real-TimeDataProcessing:High-performancecomputingallowsfortherapidanalysisof sensor data, enabling immediate responses to operational changes, such as machine failures or material shortages.

Manufacturing Area of Interest

The ARM Institute, with input from our advisory councils and the Department of Defense, has identified several Manufacturing Areas of Interest. Teams should consider these manufacturing applications as example use cases for their development and demonstration, however we encourage and expect other manufacturing areas to be part of the submissions.

·???????? Casting and Forging: Cast and forged components lie at the heart of critical commercial and defense systems, providing a vital contribution to warfighter readiness for the United States. This problem is particularly exacerbated by the nature of legacy platforms, whose designs and processes were largely conceived, defined, and stored on paper. In tandem with the pervasive challenge of workforce availability, the challenges for the United States to produce low-volume cast and forged components pose a critical and enduring issue.

·???????? Manufacturing of Hypersonics: Manufacturing hypersonic systems involve overcoming significant technical challenges due to the extreme speeds and conditions these vehicles encounter. Materials capable of withstanding high temperatures, propulsion systems, guidance mechanisms, and aerodynamic designs are critical aspects being researched and developed in this field. Because of these new materials, robotics and automation solutions are necessary to overcome manufacturing challenges in the production of components and systems in the Hypersonics sector.

·???????? Manufacturing of Energetics: Technologies related to the safe manufacturing of energetics continues to be a high priority for various ARM stakeholders. Traditional energetic material manufacturing processes are labor intensive and involve several non- linear, non-concurrent steps which add to the cycle time and logistical burden to produce in-process material as well as finished pieces. Automation of key steps can be incorporated into the manufacturing process to reduce or eliminate human-in-the-loop interactions.

·???????? Manufacturing of Garments and Other Textile Goods: There is significant market demand for apparel and other textile goods from the Defense and Commercial sectors, but the United States presently lacks the ability to compete with international manufacturers due to, among other things, high labor costs and gaps in automation. Maturation, integration, and adoption of automation and robotic technologies will be essential to secure and scale up a defense industrial base for the domestic supply chain of textile goods.

·???????? Robotic Inspection in Confined Spaces: Robotics inspection in confined spaces involves the use of advanced robotic systems to examine areas that are difficult or dangerous for human inspectors to access. Equipped with high-resolution cameras, thermal imaging, and multispectral sensors, these robots provide critical visual data for assessing structural integrity and detecting anomalies. This technology is particularly useful on submarines or ships, ensuring operational readiness and safety without endangering crew members.