Live Green Eric Frew and Sriram Sankaranarayanan University of - - PowerPoint PPT Presentation

FMCAD 2018

Live Green

Run-time Assurance for UAVs using Stochastic Modeling and Reachability Analysis

Hansol Yoon, Yi Chou, Eric Frew and Sriram Sankaranarayanan University of Colorado, Boulder FMCAD’18 Student Forum · October 31, 2018

* Unmanned Aerial Vehicles

FMCAD 2018

Motjvatjon & Objectjve

• UAVs are increasingly common in crowded urban spaces

with risks to life and property due to disturbances such as wind.

• Forecasts future UAVs positions to predict and avoid collisions.
• Quantify risk of collisions with fixed obstacles.

source: Global Trade Magazine source: ABC news

FMCAD 2018

Flight Data Flight Data UAV Platform UAV Platform Reachability Analysis Reachability Analysis Model Inference Model Inference

Model Data Warnings, Control (future work)

Predictjve Monitoring Framework

Core Position Model Future Velocity Forecast sub-model

Waypoint Guidance Algorithm Waypoint Guidance Algorithm

Acceleration Future Waypoints

Deviation/Disturbance sub-model <Architecture for the framework>

FMCAD 2018

Collision Predictjon Results

Square: Obstacle Red line: Ground truth trajectory Blue lines: Predicted trajectories

■ Evaluation on Talon UAV Flight Test Data

Real Prediction SAFE COLLISION SAFE 94 COLLISION 5 94 NOT SURE 1 6

T e s t c

• n

d i t i

• n

s :

1 . P r

• b

a b i l i t y

• f

c

• l

l i s i

• n

> = . 4 2 . P r e d i c t i

• n

t i m e h

• r

i z

• n

: 2 5 s e c s 3 . T h e d i s t a n c e

• f

t h e c e n t e r

• f

a n

• b

s t a c l e

• 2

5 m f

• r

S A F E t e s t s

• m

f

• r

C O L L I S I O N t e s t s 4 . Wi n d : 3 m / s