Proposal Summary PI is currently being funded through the NSF IUCRC grant: IIP-0856090. This proposal seeks REU supplemental funding for one undergraduate student to work on a supporting topic. We are building a system as a motivator for parallel architectures, compilers, and applications. The system consists of two automated robots. First, the sentry turret is an intelligent turret, which constantly watches for intruders in its protection zone using Microsoft Kinect, and shoots darts at the intruder. The second is a collision-avoiding car, which has to pass through the space guarded by the turret to reach its target. As the car goes through the protection zone of the turret, the turret shoots at the car, and the car intends to escape from the darts coming from the turret. The collision-avoiding car uses Kinect for RGB and Depth, Open CV for Object Detection, ArduPilot Mega to Control Servos on Traxxas Car. It runs the image processing on Ion Atom Motherboard. Development is on Microsoft Robotics Developer Studio, and Kinect SDK. The sentry turret uses Kinect for RGB and Depth, OpenCV for Object detection/tracking, Arduino to control Servos for Turret. Develoment is on Ubuntu 10.04& Libfreenect.
Proposed Student Work The proposed project has a strong experimental component which makes it an excellent candidate for introducing undergraduate students to research. Since the duration of the project is one year, the scope of the students work is going to be divided into 3 parts each corresponding to an academic period: Fall 2012, Spring 2013 and Summer 2013. The students will have to accomplish tasks of increasing complexity and independence from direct supervision. One student is going to work on developing and analyzing SCPS models. The student will be first introduced to modeling CPS and, in particular, SCPS, in order to contribute towards the SCPS benchmark library. The student will select a deterministic model from the existing benchmark library of CPS that we provide with our toolbox S-TaLiRo . In collaboration with the PI Fainekos, the graduate student working on the project, Bardh Hoxha, and engineers from Toyota, which is an IAB member company, the undergraduate student will investigate and reason about the types of stochastic behavior that can be introduced into the model. The main goal will be to develop a stochastic model from the automotive domain which is representative of the more complex models used in practice. We envision that the collaboration with our industry partner will motivate and engage the student in the modeling work. During the spring semester, the undergraduate student will be involved in formulating formal functional requirements for the model developed during the fall semester. Moreover, she/he will be trained in conducting numerical experiments using the extension of S-TaLiRo for stochastic systems. The training will be focused on formulating meaningful experiments for the experimental evaluation  of stochastic search algorithms for the CPS falsification problem. Finally, during the summer period, the undergraduate student will repeat the process on a new benchmark problem, but now with greater independence on her/his work. The other student will work on extending S-TaLiRo for testing/verifying hardware-in-the-loop (HIL) systems. S-TaLiRo is built as a Matlab/Simulink toolbox. Simulink offers the capability of performing HIL simulations. Therefore, extending S-TaLiRo to handling HIL systems will first involve the student learning the S-TaLiRo software architecture and software engineering development methods. This training and work will be performed during Fall 2012. The rest of the time will be invested on applying S-TaLiRo on the multi-UAV experimental platform of the CPSLab run by PI Fainekos. The goal will be to use S-TaliRo to find the worst case robustness scenario of an aircraft collision avoidance protocol. The multi-UAV platform consists of indoor quadrotors with a motion tracking system and it is being developed by graduate students at the CPSLab. The undergraduate student will have the opportunity to interact with them on setting up the experiments. The benefit of using UAVs for the experiments is that after each test scenario is performed the UAVs can be automatically initialized to the next position for the next experiment.
The objective of this project is to develop techniques to optimally implement dynamically reconfigurable analog proportional-integral-derivative (PID) controller on field programmable analog array (FPAA) platforms. The project will also demonstrate the advantages of FPAA PID controllers in automotive applications. Analog PID controllers can directly perform PID computation on analog data and, thus, do not need data converter circuits that are not dispensable in digital PID circuits. This makes analog PID controllers extremely appealing in applications (e.g. automotive applications) that demand low cost, high reliability, and fast responding time. However, due to the lack of programmability in analog circuits, current analog PID controllers do not have the flexibility to program coefficients used PID computation. This is the major factor that limits the use of analog PID controllers in many automotive applications. In this project, we investigate how to effectively implement PID control mechanisms, which are used in automotive throttle and suspension control, on Anadigm dynamically reconfigurable FPAAs. Since both circuit structures and analog component values can be dynamically programmed for FPAA circuits, the implemented PID controller will have same flexibility as digital PID circuits on adjusting the coefficients in PID control. In addition, we will develop techniques to perform self-testing (both off-line and on-line) for the analog PID circuits to improve the reliability of the FPAA PID controller.
|Effective start/end date||3/1/09 → 2/28/15|
- National Science Foundation (NSF): $344,303.00