Zachary Sunberg
Professor
CU Boulder
Waymo Image By Dllu - Own work, CC BY-SA 4.0, https://commons.wikimedia.org/w/index.php?curid=64517567
Tweet by Nitin Gupta
29 April 2018
https://twitter.com/nitguptaa/status/990683818825736192
Two Objectives for Autonomy
Minimize resource use
(especially time)
Minimize the risk of harm to oneself and others
Safety often opposes Efficiency
Homework on a Saturday???
Do Homework
Watch Youtube
Watch Super Bowl!
Do Homework
Turn In Homework
Homework Incomplete
Watch Super Bowl!
Turn In Homework
Saturday
Sunday
Turn Left
Go Straight
Turn Right
Crash
Destination
L
Straight
R
L
Straight
R
L
Straight
R
L
Straight
R
Partially Observable Markov Decision Process
(POMDP)
1) ACAS
2) Orbital Object Tracking
4) Asteroid Navigation
3) Dual Control
ACAS X
Trusted UAV
Collision Avoidance
[Sunberg, 2016]
[Kochenderfer, 2011]
\(\mathcal{S}\): Information space for all objects
\(\mathcal{A}\): Which objects to measure
\(R\): - Entropy
Approximately 20,000 objects >10cm in orbit
1) ACAS
2) Orbital Object Tracking
4) Asteroid Navigation
3) Dual Control
State \(x\) Parameters \(\theta\)
\(s = (x, \theta)\) \(o = x + v\)
POMDP solution automatically balances exploration and exploitation
1) ACAS
2) Orbital Object Tracking
4) Asteroid Navigation
3) Dual Control
Dynamics: Complex gravity field, regolith
State: Vehicle state, local landscape
Sensor: Star tracker?, camera?, accelerometer?
Action: Hopping actuator
[Hockman, 2017]
1) ACAS
2) Orbital Object Tracking
4) Asteroid Navigation
3) Dual Control
[Diallo, MS Thesis, 2024]
\[f(x) = \sigma(W_2 \, \sigma(W_1 x + b_1) + b_2)\]
Image
\((x)\)
"Human here"
[Diallo, MS Thesis, 2024]
\[f(x) = \sigma(W_2 \, \sigma(W_1 x + b_1) + b_2)\]
Image
\((x)\)
Baseline
Our POMDP Planner
[Ray, Laouar, Sunberg, & Ahmed, ICRA 2023]