Skip Navigation

Research Tasks

Swarm Autonomy

The goal of this effort is to develop swarm (multi/distributed) autonomy capabilities to achieve better mission performance, robustness and flexibility/reconfiguration at lower cost compared to the single spacecraft. One objective has been to develop physics-based swarm autonomous GNC algorithms for precision navigation and distributed estimation of 3D target shape, combined with optimal planning and reconfiguration.

In the technical area of swarm motion planning, we developed a multi-agent trajectory planning framework that generates trajectories for optimal estimation (localization, relative navigation, and target shape/pose estimation) while optimizing the mission cost function. The main results have been: a high-level planner was developed that generates and assigns waypoints for multi-agent trajectories near optimally; a low-level planner incorporating Sequential Convex Programming (SCP) scalable to (~100) waypoints (~10) agent size; computation within a few seconds (less than 5) was developed and tested for dynamic feasibility.

In the technical area of swarm localization and 3D mapping we analyzed and implemented on board image-based localization and 3D reconstruction for use with navigation and coverage determination. The main results have been: an analysis of different feature detection algorithms as applied to asteroids; an analysis and implementation of a real-time localization and 3D reconstruction algorithm on Jetson TX2 to work in the Caltech CAST arena; implementation of a mixed 2D-3D mapping algorithm for higher resilience in optical tracking; and identification of a new and promising swarm based localization technique.

The hardware development and system integration and testing focused on building two customized quadrotors and five 6-DOF and 3-DOF spacecraft simulators. A common avionics system was developed for both quadrotors and spacecraft simulators. Flight tests of quadrotors were carried out at CAST using an initial set of trajectories from the planning work and sample image datasets were collected from live flights to test the 3D reconstruction algorithm.
Point of Contact: Adrian Stoica
Sponsored By: JPL Internal Research Programs funded