William Pope
M.S. Thesis Defense
14 Nov 2022
Holonomic | Nonholonomic | |
Speed | Bai, H. | Bai, H. |
Speed/Steering | Gupta, H. |
Holonomic planning
Nonholonomic planning
steering limited based on velocity
Contributions
Overview
Goal: create UB/LB rollout policies that meet requirements
HJB PDE
Policy | Runtime [μs] |
---|---|
HJB | 18.064 |
Reactive-HJB | 6.083 |
Approx Reactive-HJB | 1.915 |
Lower Bound Policy | Average Time-to-Goal [s] | Average Node Rate [1/s] |
---|---|---|
Reactive-HJB | 17.9 | 759.4 |
Approx Reactive-HJB | 16.5 | 820.0 |
N = 10
Goal: add safety mechanism to handle low probability collisions and finite braking ability
Ye, N., et al
[Chen, M.]
Humans in Environment | Average Time-to-Goal [s] | Interventions / Actions | Collisions / Trials |
---|---|---|---|
10 (off) | 15.4 | n/a | 0.1 |
10 (on) | 23.0 | 0.207 | 0 |
20 (off) | 17.2 | n/a | 0.2 |
20 (on) | 31.3 | 0.280 | 0 |
N = 10
Hardware stack
Intel NUC i5
Online decision loop
\(f_{loop} = 2 \text{ Hz}\)
Future work
William Pope