Sarah M. Loos and André Platzer Computer Science Department Carnegie Mellon University October, 2011
Ultimately 1/12 Simplifying Assumptions Vehicles have positive - - PowerPoint PPT Presentation
Ultimately 1/12 Simplifying Assumptions Vehicles have positive - - PowerPoint PPT Presentation
Safe Intersections: At the Crossing of Hybrid Systems and Verification Sarah M. Loos and Andr Platzer Computer Science Department Carnegie Mellon University October, 2011 Ultimately 1/12 Simplifying Assumptions Vehicles have
Ultimately…
1/12
Simplifying Assumptions
2/12
- Vehicles have positive velocity
- Accurate sensing
- Instantaneous braking and acceleration
- Time synchronization
- Delay for sensor updates is bounded
- Straight lane dynamics
- Cars represented as points, lanes as lines
Previous Work: Highway Control
3/12
- Verified multilane highway system
- Arbitrary number of cars
- Arbitrary number of lanes
- Proof of safety for distributed control
built from two-car “building blocks.”
Intersection Building Blocks
4/12
Intersection Building Blocks
4/12
Intersection Building Blocks
4/12
Intersection Building Blocks
4/12
Intersection Building Blocks
4/12
Intersection Building Blocks
4/12
Intersection Building Blocks
4/12
Intersection Building Blocks
4/12
This is similar to a merge on the highway.
T-Intersection Building Block
5/12
Intersection Building Blocks
6/12
Intersection Building Blocks
6/12
Intersection Building Blocks
6/12
Intersection Building Blocks
6/12
Straight Lane Building Block
7/12
Di Different ntial Dyna l Dynami mic L Logic*
*
Initial Conditions → [Model] Requirements
*The
he s sho hort v version. n.
8/12
Initial Conditions → [Model] Requirements
Di Different ntial Dyna l Dynami mic L Logic
8/12
Initial Conditions → [Model] Requirements
logical formula logical formula
Di Different ntial Dyna l Dynami mic L Logic
8/12
Initial Conditions → [Model] Requirements
logical formula logical formula
Di Different ntial Dyna l Dynami mic L Logic
8/12
Initial Conditions → [Model] Requirements
logical formula logical formula
Di Different ntial Dyna l Dynami mic L Logic
8/12
Initial Conditions → [Model] Requirements
logical formula logical formula hybrid program
Di Different ntial Dyna l Dynami mic L Logic
8/12
Initial Conditions → [Model] Requirements
logical formula logical formula hybrid program discrete control continuous dynamics
Di Different ntial Dyna l Dynami mic L Logic
8/12
logical formula logical formula hybrid program
→ [(ctrl;dyn)*]
discrete control continuous dynamics
Di Different ntial Dyna l Dynami mic L Logic
8/12
logical formula logical formula hybrid program
→ [(ctrl; x’= v; v’= a)*]
discrete control continuous dynamics
Di Different ntial Dyna l Dynami mic L Logic
8/12
Single Lane Stoplight
Initial Conditions → [Model] Requirements
To Prove:
9/12
Single Lane Stoplight
Initial Conditions → [Model] Requirements
To Prove:
9/12
Single Lane Stoplight
Initial Conditions → [Model] Requirements
To Prove:
9/12
Single Lane Stoplight
Initial Conditions → [Model] Requirements
To Prove:
9/12
Single Lane Stoplight
Initial Conditions → [Model] Requirements
To Prove:
9/12
Single Lane Stoplight
Initial Conditions → [Model] Requirements
To Prove:
9/12
Single Lane Stoplight
Initial Conditions → [Model] Requirements
To Prove:
9/12
Single Lane Stoplight
Initial Conditions → [Model] Requirements
To Prove:
9/12
Single Lane Stoplight
Initial Conditions → [Model] Requirements
To Prove:
9/12
Single Lane Stoplight
Initial Conditions → [Model] Requirements
To Prove:
9/12
Single Lane Stoplight
Initial Conditions → [Model] Requirements
To Prove:
9/12
Single Lane Stoplight
Initial Conditions → [Model] Requirements
To Prove:
9/12
Single Lane Stoplight
Initial Conditions → [Model] Requirements
To Prove:
9/12
Single Lane Stoplight
Initial Conditions → [Model] Requirements
To Prove:
9/12
Single Lane Stoplight
Initial Conditions → [Model] Requirements
To Prove:
h
✔
9/12
Intersection
Initial Conditions → [Model] Requirements
To Prove:
10/12
Intersection
Initial Conditions → [Model] Requirements
To Prove: Cars can stop initially
10/12
Intersection
Initial Conditions → [Model] Requirements
To Prove: Cars can stop initially No collision
10/12
Intersection
Initial Conditions → [Model] Requirements
To Prove:
Cars can stop initially
No collision
10/12
Intersection
Initial Conditions → [Model] Requirements
To Prove:
Cars can stop initially
No collision
10/12
Intersection
Initial Conditions → [Model] Requirements
To Prove:
Cars can stop initially
No collision
10/12
Intersection
Initial Conditions → [Model] Requirements
To Prove:
Cars can stop initially
No collision
10/12
Intersection
Initial Conditions → [Model] Requirements
To Prove:
Cars can stop initially
No collision
10/12
Intersection
Initial Conditions → [Model] Requirements
To Prove:
Cars can stop initially
No collision
10/12
Intersection
Initial Conditions → [Model] Requirements
To Prove:
Cars can stop initially
No collision
h
✔
10/12
Conclusions Future Work
11/12
Curved road dynamics Distributed car dynamics Combinations of merge and cross protocols Noisy and delayed sensor data Delayed braking and acceleration reaction Non-synchronized time Non-zero car lengths and lane widths
Cha halle lleng nges Solu lutions ns
Infinite, continuous, and evolving
state space, R∞
Simulation and testing only
partially prove safety
Continuous dynamics Discrete control decisions Large branching factor
Conclusions
We give a formal proof for a two-lane
intersection with one car on each lane
Semi-automated proof generation Variations in system design Demonstrated potential for formal
safety verification in car control, even when models have high branching factor
12/12
Conclusions Thank You!
12/12
Conclusions Reference
12/12 The full length paper for this research can be found here: Sarah M. Loos and André Platzer. Safe Intersections: At the Crossing of Hybrid Systems and Verification. In the 14th International IEEE Conference on Intelligent Transportation Systems, ITSC 2011, Washington, D.C., USA, Proceedings, 2011.