This paper presents a learning framework to estimate agent capabilities and task requirements for multi-robot task allocation, where the learned capabilities and requirements can be embedded as constraints in many existing optimization-based task allocation models.

[1] B. Fu, W. Smith, D. Rizzo, M. Castanier, M. Ghaffari, and K. Barton, “Learning Task Requirements and Agent Capabilities for Multi-agent Task Allocation,” arXiv preprint arXiv:2211.03286, 2022. [Under review] [PDF] [Code] [Video] [Website]

Recent advances in robotic sensing and control have enabled the application of robotic systems in various unstructured natural environments, including robotic agriculture, underwater exploration, and search and rescue in caves. In the meantime, the progress in multi-robot decision-making algorithms and the growth in computational power allow robot swarms to work together to complete missions. Despite the achievements made by these decision-making frameworks, most of them are applied to a team of homogeneous robots for one specific task and require a structured known environment. As robotic systems operate on more complex missions, it is essential to consider uncertainty in the environment and the heterogeneity of the robots and tasks.

In this project, we aim to develop a general feedback teaming framework that dynamically characterizes the heterogeneities and uncertainties in energy, time, vehicle states, and task states; and configure robust and agile teams and schedules despite the heterogeneities and uncertainties.

[1] B. Fu, W. Smith, D. Rizzo, M. Castanier, and K. Barton, “Heterogeneous vehicle routing and teaming with Gaussian distributed energy uncertainty,” in 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2020, pp. 4315-4322 [PDF] [Presentation]

[2] B. Fu, W. Smith, D. Rizzo, M. Castanier, M. Ghaffari, and K. Barton, “Robust task scheduling for heterogeneous robot teams under capability uncertainty,” IEEE Transactions on Robotics, vol. 39, no. 2, pp. 1087–1105, 2022. [PDF] [Code] [Video] [Website]

[3] Article: Assembling the Right Team for the Mission 

This work presents a centralized framework for multi-robot tour guidance in a partially known environment. A mixed-integer optimization simultaneously matches humans with the correct robot and generates the routes for the robots to maximize the coverage of requested places to visit. The scalability and optimality of the framework are demonstrated through computational evaluation (largest case tested: 50 robots, 250 humans, and 50 places). A photo-realistic simulation was developed to verify the tour guiding performance in an uncertain indoor environment.

[1] B. Fu, T. Kathuria, D. Rizzo, M. Castanier, X. J. Yang, M. Ghaffari, and K. Barton, “Simultaneous human-robot matching and routing for multi-robot tour guiding under time uncertainty,” Journal of Autonomous Vehicles and Systems, vol. 1, no. 4, p. 041005, 2021. [PDF] [Code] [Website]

State-of-the-art forward facing monocular visual-inertial odometry algorithms are often brittle in practice, especially whilst dealing with initialization and motion in directions that render the state unobservable. In such cases having a reliable complementary odometry algorithm enables robust and resilient flight. Using the common local planarity assumption, we present a fast, dense, and direct frame-to-frame visual-inertial odometry algorithm for downward facing cameras that minimizes a joint cost function involving a homography based photometric cost and an IMU regularization term. Via extensive evaluation in a variety of scenarios we demonstrate superior performance than existing state-of-the-art downward facing odometry algorithms for Micro Aerial Vehicles (MAVs).

[1] B. Fu, K. S. Shankar, N. Michael, “Rad-VIO: Rangefinder-aided downward visual-inertial odometry,” in 2019 International Conference on Robotics and Automation (ICRA). IEEE, 2019, pp. 1841-1847. [PDF]

We developed a rule-based control strategy, which achieved a 24.41% fuel consumption decrease in the Simulink-based simulation compared to the result of the original internal combustion engine vehicle. We used genetic algorithm to optimize the hyper-parameters in the control strategy and achieved an additional 1.53% fuel consumption reduction. Finally, extensive hardware-in-the-loop tests for the hybrid control unit have been conducted to prove the functionality, reliability, and robustness on real hardware.

[1] J. Hao, Z. Yu, Z. Zhao, X. Zhan, B. Fu, and P. Shen, “Development and optimization of energy management strategy for four-wheel-drive plug-in hybrid electric vehicle,” Mechatronics Journal, 2018, no. 8, pp.12-19, 30.