This document presents a sampling-based approach to spacecraft planning with safety specifications. It proposes generating collision avoidance maneuvers (CAMs) that can safely navigate a spacecraft to a positively invariant set for any failure time or mode. This simplifies the safety problem by only requiring checking if CAMs are feasible under different failure scenarios, rather than solving complex optimization problems. The approach restricts spacecraft burns and planning to actively safe nodes defined by the CAMs, allowing safety to be incorporated into sampling-based planning methods. This provides a provably correct framework for systematically encoding safety specifications into spacecraft trajectory generation.
Simulation-based Test Case Generation for Unmanned Aerial Vehicles in the Nei...Sebastiano Panichella
Here are the slides of the presentation of the paper entitled "Simulation-based Test Case Generation for Unmanned Aerial Vehicles in the Neighborhood of Real Flights". It was presented at the IEEE International Conference on Software Testing, Verification, and Validation (ICST) 2023.
The presentation concerns the ongoing research in the COSMOS H2020 project (https://www.cosmos-devops.org/), as outlined by the ICST Program (https://conf.researchr.org/program/icst-2023/program-icst-2023/?past=Show%20upcoming%20events%20only).
Here are the slides of my presentation of the paper entitled "Simulation-based Test Case Generation for Unmanned Aerial Vehicles in the Neighborhood of Real Flights". It was presented at the IEEE International Conference on Software Testing, Verification, and Validation (ICST) 2023 (https://conf.researchr.org/home/icst-2023).
The presentation concerns the ongoing research in the COSMOS H2020 project (https://www.cosmos-devops.org/).
Optimal Sensor Management Technique For An Unmanned Aerial Vehicle Tracking M...Negar Farmani
In this paper, we present an optimal sensor management technique for an Unmanned Aerial Vehicle (UAV) to autonomously geo-localize multiple mobile ground targets. The target states are continuously estimated using target locations asynchronously captured by a gimbaled camera with a limited field of view and processed with a set of Extended Kalman Filters (EKFs). The technique incorporates a Dynamic Weighted Graph (DWG) method to first group estimated targets and then determine regions with high target densities. A Model Predictive Control (MPC) method is used to compute a camera pose that minimizes the overall uncertainty of the target state estimates. The validity of the proposed technique is demonstrated using simulation results.
Simulation-based Test Case Generation for Unmanned Aerial Vehicles in the Nei...Sebastiano Panichella
Here are the slides of the presentation of the paper entitled "Simulation-based Test Case Generation for Unmanned Aerial Vehicles in the Neighborhood of Real Flights". It was presented at the IEEE International Conference on Software Testing, Verification, and Validation (ICST) 2023.
The presentation concerns the ongoing research in the COSMOS H2020 project (https://www.cosmos-devops.org/), as outlined by the ICST Program (https://conf.researchr.org/program/icst-2023/program-icst-2023/?past=Show%20upcoming%20events%20only).
Here are the slides of my presentation of the paper entitled "Simulation-based Test Case Generation for Unmanned Aerial Vehicles in the Neighborhood of Real Flights". It was presented at the IEEE International Conference on Software Testing, Verification, and Validation (ICST) 2023 (https://conf.researchr.org/home/icst-2023).
The presentation concerns the ongoing research in the COSMOS H2020 project (https://www.cosmos-devops.org/).
Optimal Sensor Management Technique For An Unmanned Aerial Vehicle Tracking M...Negar Farmani
In this paper, we present an optimal sensor management technique for an Unmanned Aerial Vehicle (UAV) to autonomously geo-localize multiple mobile ground targets. The target states are continuously estimated using target locations asynchronously captured by a gimbaled camera with a limited field of view and processed with a set of Extended Kalman Filters (EKFs). The technique incorporates a Dynamic Weighted Graph (DWG) method to first group estimated targets and then determine regions with high target densities. A Model Predictive Control (MPC) method is used to compute a camera pose that minimizes the overall uncertainty of the target state estimates. The validity of the proposed technique is demonstrated using simulation results.
TAROT2013 Testing School - Gilles Perrouin presentationHenry Muccini
TAROT 2013 9th International Summer School on Training And Research On Testing, Volterra, Italy, 9-13 July, 2013
These slides summarize Gilles Perrouin's presentation about "Feature-based Testing of SPLs: Pairwise and Beyond"
TAROT2013 Testing School - Gilles Perrouin presentationHenry Muccini
TAROT 2013 9th International Summer School on Training And Research On Testing, Volterra, Italy, 9-13 July, 2013
These slides summarize Gilles Perrouin's presentation about "Feature-based Testing of SPLs: Pairwise and Beyond"
1. A Sampling-Based Approach to Spacecraft
Autonomous Maneuvering with Safety Specifications
Joseph A. Starek∗, Brent W. Barbee†, Marco Pavone∗
∗Autonomous Systems Lab †Navigation and Mission Design Branch
Dept. of Aeronautics & Astronautics Goddard Space Flight Center
Stanford University NASA
AAS-GNC 2015
February 3rd, 2015
Supported by NASA STRO-ECF Grant #NNX12AQ43G
2. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Outline
• Autonomous Vehicle Safety
1 / 21
3. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Outline
• Autonomous Vehicle Safety
• Spacecraft Safety
1 / 21
4. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Outline
• Autonomous Vehicle Safety
• Spacecraft Safety
• Safety in CWH Dynamics
1 / 21
5. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Outline
• Autonomous Vehicle Safety
• Spacecraft Safety
• Safety in CWH Dynamics
• Numerical Experiments
1 / 21
6. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Outline
• Autonomous Vehicle Safety
• Spacecraft Safety
• Safety in CWH Dynamics
• Numerical Experiments
• Conclusions and Future Work
1 / 21
7. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
The Need for Safe Autonomy
1 / 21
8. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
The Need for Safe Autonomy
• Satellite servicing (DARPA
Phoenix Mission)
1 / 21
9. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
The Need for Safe Autonomy
• Satellite servicing (DARPA
Phoenix Mission)
• Automated rendezvous
1 / 21
10. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
The Need for Safe Autonomy
• Satellite servicing (DARPA
Phoenix Mission)
• Automated rendezvous
Key Question
How do we implement a general, automated spacecraft
planning framework with hard safety specifications?
1 / 21
11. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Original Contribution
Our work:
1. Establishes a provably-correct framework for the
systematic encoding of safety specifications into the
spacecraft trajectory generation process
2. Derives an efficient one-burn escape maneuver policy
for proximity operations near circular orbit
2 / 21
12. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Previous Work
Spacecraft rendezvous approaches with explicit
characterizations of safety:
• Kinematic path optimization [Jacobsen, Lee, et al., 2002]
• Artificial potential functions [Roger and McInnes, 2000]
• MILP formulations [Breger and How, 2008]
• Safety ellipses [Gaylor and Barbee, 2007] [Naasz, 2005]
• Motion planning [Frazzoli, 2003]
• Robust Model-Predictive Control [Carson, Ac¸ikmes¸e, et
al., 2008]
• Forced equilibria [Weiss, Baldwin, et al., 2013]
3 / 21
13. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Types of Spacecraft Rendezvous Safety
• Passive Trajectory Protection: Constrain coasting
trajectories to avoid collisions up to a given horizon time
• Active Trajectory Protection: Implement an actuated
escape maneuver to save/abort a mission
Design Choice
We emphasize active safety as it is the less-conservative
approach
4 / 21
14. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Vehicle Trajectory Safety
Definition (Trajectory Safety Problem)
For all possible failure times tfail ∈ Tfail and failure modes
Ufail(x(tfail)), we seek a sequence of admissible actions
u(τ) ∈ Ufail(x(tfail)) from x(tfail) such that the remaining
trajectory is safe.
Examples:
• Rovers/Land vehicles: Come to a complete stop
• Manipulators: Return to previous configuration,
disengage, or execute emergency plan
• UAV’s: Enter a safe loiter pattern
• Spacecraft: Less straightforward; generally require
mission-specific solutions (with human oversight)
5 / 21
15. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Vehicle Trajectory Safety
Definition (Trajectory Safety Problem)
For all possible failure times tfail ∈ Tfail and failure modes
Ufail(x(tfail)), we seek a sequence of admissible actions
u(τ) ∈ Ufail(x(tfail)) from x(tfail) such that the remaining
trajectory is safe.
6 / 21
16. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Vehicle Trajectory Safety
Definition (Trajectory Safety Problem)
For all possible failure times tfail ∈ Tfail and failure modes
Ufail(x(tfail)), we seek a sequence of admissible actions
u(τ) ∈ Ufail(x(tfail)) from x(tfail) such that the remaining
trajectory is safe.
6 / 21
17. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Vehicle Trajectory Safety
Definition (Trajectory Safety Problem)
For all possible failure times tfail ∈ Tfail and failure modes
Ufail(x(tfail)), we seek a sequence of admissible actions
u(τ) ∈ Ufail(x(tfail)) from x(tfail) such that the remaining
trajectory is safe.
Failure
6 / 21
18. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Vehicle Trajectory Safety
Definition (Trajectory Safety Problem)
For all possible failure times tfail ∈ Tfail and failure modes
Ufail(x(tfail)), we seek a sequence of admissible actions
u(τ) ∈ Ufail(x(tfail)) from x(tfail) such that the remaining
trajectory is safe.
Failure
6 / 21
19. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Vehicle Trajectory Safety
Definition (Trajectory Safety Problem)
For all possible failure times tfail ∈ Tfail and failure modes
Ufail(x(tfail)), we seek a sequence of admissible actions
u(τ) ∈ Ufail(x(tfail)) from x(tfail) such that the remaining
trajectory is safe.
Failure
6 / 21
20. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Vehicle Trajectory Safety
Definition (Trajectory Safety Problem)
For all possible failure times tfail ∈ Tfail and failure modes
Ufail(x(tfail)), we seek a sequence of admissible actions
u(τ) ∈ Ufail(x(tfail)) from x(tfail) such that the remaining
trajectory is safe.
Failure
6 / 21
21. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Challenge: Infinite-Horizon Safety
Finite-horizon safety guarantees can ultimately violate
constraints:
Failure
7 / 21
22. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Idea: Positively-Invariant Sets
Definition (Positively-Invariant Set)
A set Xinvariant is positively invariant with respect to
˙x = f (x) if and only if
x(t0) ∈ Xinvariant =⇒ x(t) ∈ Xinvariant, t ≥ t0
8 / 21
23. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Idea: Positively-Invariant Sets
Definition (Vehicle State Safety)
A state is safe if and only if there exists, under all failure
conditions, a safe, dynamically-feasible trajectory that
navigates the vehicle to a safe, stable positively-invariant set.
Failure
9 / 21
24. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Finite-Time Trajectory Safety
minimize
tf,x(t),u(t)
J(x, u, t)
subject to ˙x(t) = f (x(t), u(t), t) (Dynamics)
x(t0) = x0 (Initial Condition)
x(tf) ∈ Xinvariant (Invariant Termination)
u(t) ∈ Ufail(x0) (Control Admissibility)
gi (x, u) ≤ 0, i = [1, . . . , p] (Inequality Constraints)
hj(x, u) = 0, j = [1, . . . , q] (Equality Constraints)
10 / 21
25. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Challenge: Solving the Finite-Time Safety
Problem under Failures
For a K-fault tolerant spacecraft with N control components
(thrusters, momentum wheels, CMG’s, etc), this yields:
Nfail =
K
k=0
N
k
=
K
k=0
N!
k!(N − k)!
total optimization problems (one for each Ufail) for each
failure time tfail.
11 / 21
26. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Idea: Simplify the Finite-Time Safety Problem
Theorem (Sufficient Fault-Tolerant Active Safety)
1. From each x(tfail), prescribe a Collision-Avoidance
Maneuver ΠCAM(x) that gives a horizon T and escape
sequence u that satisfies x(T) ∈ Xinvariant and u(τ) ⊂ U
for all tfail ≤ τ ≤ T.
2. For each failure mode Ufail(x(tfail)) ⊂ U(x(tfail)) up to
tolerance K, check if u = ΠCAM(x(tfail)) ⊂ Ufail(x(tfail)).
12 / 21
27. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Idea: Simplify the Finite-Time Safety Problem
Theorem (Sufficient Fault-Tolerant Active Safety)
1. From each x(tfail), prescribe a Collision-Avoidance
Maneuver ΠCAM(x) that gives a horizon T and escape
sequence u that satisfies x(T) ∈ Xinvariant and u(τ) ⊂ U
for all tfail ≤ τ ≤ T.
2. For each failure mode Ufail(x(tfail)) ⊂ U(x(tfail)) up to
tolerance K, check if u = ΠCAM(x(tfail)) ⊂ Ufail(x(tfail)).
Key Simplifications
Removes decision variables u, reducing to:
• a test of escape control feasibility under failure(s)
• numerical integration for satisfaction of dynamics
• an a posteriori check of constraints gi and hj
12 / 21
28. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Safe Sampling-Based Spacecraft Planning
Solution is in exact form required for sampling-based motion
planning.
Restrict burns to nodes
Actively-safe Sampling-based
Spacecraft Planning
Restrict planning to
actively-safe nodes
Incorporating Safety Constraints:
• Add CAM policy generation to sampling algorithm
• Include CAM-trajectory collision-checking in tests of
sample feasibility
13 / 21
29. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Example: CAM Policy Design
Using CWH Set Invariance for CAMs
KOZ
Circularization RIC
Reference Line
Chaser
14 / 21
30. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Example: CAM Policy Design
Using CWH Set Invariance for CAMs
KOZ
Circularization RIC
Reference Line
Chaser
Circular Clohessy-Wiltshire-Hill (CWH) CAM policy:
1. Coast from x(t) to some new T > t such that x(T−) lies
at a position in Xinvariant.
2. Circularize the orbit at x(T) such that x(T+) ∈ Xinvariant
3. Coast along the new orbit (horizontal drift along the
in-track axis) in Xinvariant
14 / 21
31. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Example: CAM Policy Design
Choosing the Circularization Time, T
CWH Finite-Time Safety Problem:
Given: x(t), u(τ) = 0, t ≤ τ < T
minimize
T
∆v2
circ(T)
subject to ˙x(τ) = f (x(τ), 0, τ) (Dynamics)
x(τ) ∈ XKOZ (KOZ Avoidance)
x(T+
) ∈ Xinvariant (Invariant Termination)
Key Result
Can be reduced to an analytical expression that is solvable in
milliseconds
15 / 21
32. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Scenario
• Simulates an automated approach to LandSat-7 (e.g., for
servicing) between pre-specified waypoints
• Calls on the Fast Marching Tree (FMT∗) algorithm for
implementation
16 / 21
33. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Scenario
• Simulates an automated approach to LandSat-7 (e.g., for
servicing) between pre-specified waypoints
• Calls on the Fast Marching Tree (FMT∗) algorithm for
implementation
Assumptions:
• Begins at insertion into a coplanar
circular orbit sufficiently close to
the target
• The target is nadir-pointing
• The chaser is nominally
nadir-pointing, or executes a “turn
and burn” along CAMs
16 / 21
34. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Scenario
• Simulates an automated approach to LandSat-7 (e.g., for
servicing) between pre-specified waypoints
• Calls on the Fast Marching Tree (FMT∗) algorithm for
implementation
Constraints:
• Plume impingement: No exhaust
plume impingement
• Collision avoidance: Clearance of
an elliptic Keep-Out Zone (KOZ)
• Target communication: Target
comm lobe avoidance
• Safety: Two-fault tolerance to
stuck-off failures
16 / 21
35. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Motion Planning Problem
Motion planning query:
17 / 21
36. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Motion Plan Comparison
Motion planning solutions:
18 / 21
37. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Motion Plan Comparison
Motion planning solutions:
18 / 21
38. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Success Rate Comparison
Success comparison as a function of thruster failure
probability, computed over 50 trials:
0 0.02 0.04 0.06 0.08 0.1
0
0.2
0.4
0.6
0.8
1
pfailure
SuccessRate
No Safety
Active Safety
19 / 21
39. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Success Rate Comparison
Success comparison as a function of thruster failure
probability, computed over 50 trials:
0 1 2 3 4 5 6 7 8
0
24
48
72
96
120
Number of Failed Thrusters
Trials
Success
Failure
0 1 2 3 4 5 6 7 8
0
24
48
72
96
120
Number of Failed Thrusters
Trials
Success
Success (CAM)
Failure
19 / 21
40. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Conclusions
Key Ideas
1. Use termination constraints inside safe, stable,
positively-invariant sets for infinite-horizon maneuver
safety
2. Embed invariant-set constraints into sampling-based
algorithms for safety-constrained planning
Synopsis
• Demonstrated the idea for failure-tolerant circular CWH
planning
• CAM policies can be precomputed offline for more
efficient online computation
20 / 21
41. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Future Work
Future Goals
• Extend to thruster stuck-on
and mis-allocation failures
• Account for localization
uncertainty
• Apply these notions to
small-body proximity
operations
21 / 21
42. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Autonomous
Vehicle Safety
Spacecraft Safety
Active Safety with
Positively-Invariant Set
Constraints
CWH CAM Policy Design
Numerical
Experiments
Conclusions
Future Goals
Thank you!
Joseph A. Starek, Brent W. Barbee, and Marco
Pavone
Aeronautics & Astronautics Navigation and Mission Design
Stanford University NASA GSFC
jstarek@stanford.edu
21 / 21
44. Sampling-Based
Spacecraft Safety
J. Starek, B.
Barbee, M. Pavone
Dynamics
Sampling-Based
Motion Planning
Optimal Motion Planning
FMT∗
Generalized Mover’s Problem
Definition (Optimal Motion Planning Problem)
Given X, Xobs, Xfree, and J, find an action trajectory
u : [0, T] → U yielding a feasible path x(t) ∈ Xfree over
time horizon t ∈ [0, T], which reaches the goal region
x(T) ∈ Xgoal and minimizes the cost functional
J = T
0 c(x(t), u(t)) dt.
Characteristics:
• PSPACE-hard (and therefore NP-hard)
• Requires kinodynamic motion planning
• Almost certainly requires approximate algorithms, tailored
to the particular application
2 / 3