Multi-Robot Systems Projects
Fast, Real-time, Mapless Multi-Robots Navigation Across Multiple Environments
In this project, we are investigating techniques for autonomous and rapid navigation of a mobile robot in an environment with complex obstacles, without having an a priori map of the environment. In contrast to using sampling-based robot path planners, our proposed techniques rely on machine learning and AI-based decision making under uncertainty to enable robots to reuse navigation information from past experiences for fast, real-time path planning with collision avoidance.
- O. Saha. P. Dasgupta, "Real-Time Robot Path Planning Around Complex Obstacle Patterns through Learning and Transferring Options," IEEE International Conference on Robot Systems and Competitions (ICARSC) , 2017 (Best paper award). [pdf] (on Researchgate)
- O. Saha. P. Dasgupta, , "Improved Reward Estimation for Efficient Robot Navigation Using Inverse Reinforcement Learning," 11th NASA/ESA Conference on Adaptive Hardware and Systems (AHS), Pasadena, CA, 2017.
- O. Saha, P. Dasgupta, "Real-time Robot Path Planning Using Experience Learning From Common Obstacle Patterns: (Extended Abstract)", International Conference on Autonomous Agents and MultiAgent Systems (AAMAS), 2016, pp. 1339-1340. (ACM DL Link)
- O. Saha, P. Dasgupta, "Fast Path Planning using Experience Learning from Obstacle Patterns," AAAI Spring Symposium 2016 - Multiagent Learning for the Real World, 2016. [AAAI pdf link]
Multi-robot Information-driven Exploration under Communication Constraints
The objective of this project is to enable a team of robots to autonomously explore and collection information about a spatially distributed physical phenomenon (e.g., air quality, temperature gradient, soil moisture, etc.) without having prior knowledge about the location of the phenomenon inside the environment. We are addressing this problem in the context of limited or constrained communication between multiple robots, so that they can intelligently plan locations for sample collection while avoiding path conflicts and repeated exploration of the same regions.