Multi-Agent Ground Autonomous Systems
Abstract
We request support for purchasing essential instrumentation to develop next generation technologies for ground autonomy using multiple robot agents in complex indoor and outdoor scenarios. Our long-term goals are to develop legged robotic platforms with enhanced capability to autonomously navigate in heterogeneous terrains including tall grass, low brush, as well as all kinds of buildings that are navigable by humans. In addition to develop autonomous capabilities using multi-modal sensors and AI algorithms, we would also develop new techniques to coordinate the navigation of such robotic platforms for search, surveillance or warfighting missions. The underlying research will lay the scientific foundation of a new area related to ?multi-agent ground autonomous systems?, which will combine techniques from multi-agent simulation, computer vision, physics-based simulation, robot navigation and machine learning. We request support for multiple legged platforms with varying capabilities to test and evaluate the performance of our methods in different environments. This equipment will also support the ongoing research under the ARL Cooperative Agreement (ArtIAMAS) and will be used to evaluate our new methods at ARL?s Robotics Research Collaboration Campus (R2C2). The overall research results supported by this equipment will be released in the public domain. 1. Autonomous Ground Robot Navigation in Complex Scenarios 2. Perception Based Teaming and Aerial Video Recognition 3. Perception for Maneuvering Legged Robots in Dense Unstructured Terrains 4. NERF-Based Synthetic Data Generator for Aerial Video Analysis 5. Developing Novel Risk Sensitive Maneuver Tradeoffs for a Team of Ground Vehicles 6. Efficient Posterior Sampling Based Multi-Agent Reinforcement Learning via Stein?s Method for Next-generation Team of Combat Vehicles In terms of equipment support, we propose to state of the art legged platforms to develop next generation systems for multi-agent ground autonomy. More specifically, we request the latest versions of the Spot legged Platform (Enterprise Package) with the Enterprise Upgrade. Moreover, we would also kike to order Spot Enterprise Package with Spot Arm Upgrade with the Battery and Charger, as well as other accessories. Overall, we are requesting support for three new Spot legged platforms with different capabilities and configuration to evaluate our methods for multi-agent ground autonomy. In order to train and test the novel machine learning methods, we also request support for two high-end Dell workstations with 64 CPUs cores and two high-end NVIDIA Dual A6000 GPUs with high memory (256GB) and disk space (16TB). Besides multi-agent autonomous system in complex indoor and outdoor scenes, the techniques developed using the proposed equipment will also be useful for ongoing research projects on human-robot interactions, computer vision, AI and physics-based simulation.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 16, 2023
- Source ID
- W911NF2310352
Entities
People
- Dinesh Manocha
Organizations
- Army Contracting Command
- United States Army
- University of Maryland