Abstract
We consider the problem of autonomous navigation using limited information from a remote sensor network. Because the remote sensors are power and bandwidth limited, we use event-triggered (ET) estimation to manage communication costs. We introduce a fast and efficient sampling-based planner which computes motion plans coupled with ET communication strategies that minimize communication costs, while satisfying constraints on the probability of reaching the goal region and the point-wise probability of collision. We derive a novel method for offline propagation of the expected state distribution, and corresponding bounds on this distribution. These bounds are used to evaluate the chance constraints in the algorithm. Case studies establish the validity of our approach, demonstrating fast computation of optimal plans.
Introduction
Autonomous navigation in remote or resource-constrained environments, such as lunar exploration, often requires robots to communicate with external sensor networks. However, constant communication is costly, draining batteries and consuming limited bandwidth. To address this, Event-Triggered (ET) Estimation is used, where the robot only communicates when its internal state estimate becomes sufficiently uncertain—specifically, when the difference between predicted and actual measurements exceeds a threshold $\delta$.
The core challenge lies in planning: how can a robot simultaneously choose a path and a communication strategy? High thresholds save power but increase the risk of collision, while low thresholds ensure safety at the cost of energy. Existing methods to solve this trade-off are often computationally prohibitive for real-time onboard use. This paper introduces ET-GBT (Event-Triggered Gaussian Belief Trees), a fast sampling-based motion planner that balances robot safety with communication efficiency.
The ET-GBT Algorithm
ET-GBT builds on the framework of Gaussian Belief Trees and the Stable Sparse RRT (SST) planner. The algorithm is underpinned by three technical pillars:
- Offline Belief Bounding: Since sensor firing times are unknown during planning, the authors derived a novel recursive method to calculate an upper bound on state uncertainty (covariance) for any given threshold $\delta$ before the robot begins its mission.
- Joint Sampling of Paths and Thresholds: The planner samples both spatial configurations and the triggering threshold $\delta$. This allows the robot to dynamically "tighten" communication in narrow passages to shrink its uncertainty and "relax" it in open spaces to conserve resources.
- Fast Safety Verification: To maintain high planning speeds, uncertainty is represented as an $n$-dimensional sphere. If this "safety bubble" does not intersect an obstacle, the path is guaranteed to meet safety constraints within a specified probability.
Experimental Evaluation
The algorithm was tested in two primary simulated environments:
- Simple 2D System: A point-mass robot navigating a narrow corridor.
- 2nd-Order Unicycle: A complex kinematic model tested in a field of 15 randomized circular obstacles.
Key Results and Insights
Evaluation of ET-GBT demonstrated significant improvements over existing state-of-the-art methods:
- Massive Speedup: ET-GBT produced optimal plans in just 0.5 seconds, whereas previous Pareto-optimal methods required hours of computation.
- Safety Validation: In 3,000 Monte Carlo simulations, the algorithm achieved zero collisions, confirming that the conservative uncertainty bounds successfully respect safety constraints.
- Asymptotic Optimality: The planner demonstrates a consistent decrease in communication costs as more computation time is allocated, allowing it to converge toward an optimal solution.
- Dynamic Resource Management: The robot naturally learns to treat communication as a resource, decreasing the $\delta$ threshold as it enters high-risk areas and increasing it once it reaches open ground.
Conclusion
ET-GBT successfully bridges the gap between high-level communication strategy and low-level motion planning. While the current use of spherical uncertainty bounds introduces some conservatism, it enables the orders-of-magnitude speedup required for real-time applications. Future work will focus on developing tighter uncertainty bounds and extending the framework to more complex, non-linear robotic systems.
Summary updated