Cyberphysical Systems Lab

An overview of the system

An introduction and overview description of the system as well as a demo of some of the capabilities on the testbed. In particular, it focuses on the concept of ‘local temporal autonomy’ and demonstrates how the vehicles deal with communication, localization or other failures and how it recovers from them.

Tackling Autonomous Vehicle Cybersecurity Issues

This video illustrates that the dynamic watermarking method successfully detects and responds to attacks on the position sensors which can otherwise cause collisions on a laboratory autonomous transportation system

Seven Cars Scheduled in a City Traffic Scenario

This is a simulation of traffic flow through a grid traffic network. Individual cars are controlled by their own controller, however they get information on routes from a central traffic scheduler. The routes and schedules are updated periodically in order to account for new cars joining the system, cars behaving erratically or new goals. The result is a successful demonstration of a network of sensors and actuators controlled over a network of computation nodes and communication links. It does not illustrate good parking ability though and motivated the development of a collision avoidance system. (A map of the city grid is available here).

Eight Cars Scheduled in a City Traffic Scenario

A similar demonstration as the one above except with 8 cars.

Automatic Parking

Shows a car moving from one bin, through the traffic system to another bin and successfully parking. (A map of the city grid is available here).

Cars Moving in a Figure Eight

A nice illustration system operation and performance.

Principle of Safety Preserving Security Overrides

This demonstration illustrates the principle of safety preserving security overrides. This principle states that, in control systems, low-level safety features such as collision avoidance should not be overridden by higher level security overrides in general.

Collision Avoidance Implementation

This is an overview of the performance of the collision avoidance algorithm that we have implemented on the testbed. An illustration of the original problem as well as several different scenarios are presented. (This effectively eliminates situations like the one on the left!)

Component Restart Using Middleware

Illustrates the ability of the system to handle the restart of a component. During normal operation a controller for an individual car is halted and then restarted with no noticeable effect on the system performance.

Component Migration Using Middleware

Component failure of communication errors may require a component to be moved from one node to another. This video is an example of how migration has been developed and implemented in the middleware.

Component Upgrade Using Middleware

Since components can be halted and restarted without affecting the system performance there is no reason that we could not halt a particular component and start a different (perhaps better) one in its place. This video illustrates the replacement of car controllers with no adverse affects on the system.

Car Going on a Gun Shot

In collaboration with some other CSL researchers, the testbed was used to illustrate a target localization problem. Several small ‘motes’ were used to identify the location of a small gunshot. This information was passed to a controller on the testbed which computes a route to get a car to the location.



Things don’t always go as planned – and here is an example from before collision avoidance was implemented correctly…just to keep us honest!

Experimental Results Involving Delay

This video illustrates how performance is degraded in the presence of delays. The car attempting to follow a rectangular trajectory. However, the feedback is delayed by 700ms and the video illustrates the effect of known vs. unknown delay.