A UAV Test and Development Environment Based on Dynamic System Reconfiguration Osamah A. Rawashdeh, Garrett D. Chandler, and James E. Lumpp, Jr. Department of Electrical and Computer Engineering University of Kentucky Lexington, KY
Outline • Motivation/Background • Design Framework • Runtime Behavior • UAV Test and Development Environment 2/11 Intelligent Dependable Embedded Architectures Lab University of Kentucky
UAV Research at UK • BIG BLUE: Baseline Inflatable-wing Glider, Balloon- Launched Unmanned Experiment. • Ongoing project at UK to developing a test bed for Mars airplane technology. • BIG BLUE is funded by NASA and KSGC • ~ 40 students involved per year. 3/11 Intelligent Dependable Embedded Architectures Lab University of Kentucky
Framework • Software is developed in a modular fashion. • Software modules can have several implementations with different resource requirements and output qualities. • Dependencies among modules are captured in dependency graphs (DGs). • Modules are scheduled on an interconnected set of processing resources. 4/11 Intelligent Dependable Embedded Architectures Lab University of Kentucky
Framework (cont.) • Fault detection: – By application code – Heartbeat messages – OS detected violations • A system manager tracks status of hardware and software resources. • Fault handling: system is dynamically reconfigured by deploying a new mapping of software modules to hardware resources. 5/11 Intelligent Dependable Embedded Architectures Lab University of Kentucky
System Architecture – System Manager • Tracks status of System resources Processing Manager Bus • Finds and deploys Element configurations – Processing Elements Processing Processing Processing • Host I/O hardware Element Element Element • Real-time OS schedules modules Actuator Actuator Actuator – Communication Bus Sensor Sensor Sensor • CAN 2.0 standard • Control messages • Data transfer – Sensors and Actuators 6/11 Intelligent Dependable Embedded Architectures Lab University of Kentucky
Dependency Graphs • DGs show the flow of information Software module from sensors to actuators. • DG nodes: Software module V1 – Software modules Software module Executable code schedulable on a V2 processing element. Software module – Data variables V3 Inputs and outputs of software modules. m State variables are local to a software Data Variable module. – I/O devices Input Device Interface to the environment. Output Device 7/11 Intelligent Dependable Embedded Architectures Lab University of Kentucky
Data Requirements a k/n x b • Dependency symbols: c – “k-out-of-n” gates: n > 0, a 0 ≤ k ≤ n. b x c – “XOR”: only one input required. x – “DEMUX”: for fanning out. a y – “AND”: all input required. z • Quality values are associated a x b with variables. c Sensor Computing Actuator Actuator Sensor Actuator Sensor reading Module setting Driver Driver 8/11 Intelligent Dependable Embedded Architectures Lab University of Kentucky
Example Graph Release cmd High quality GPS GPS Altitude GPS altitude driver altitude (Quality1) 0/2 estimator Low quality Abs. pressure Sensor Absolute Altitude altitude sensor pressure (Quality2) driver estimator Chute Chute Chute Chute Release logic release driver Solenoid 9/11 Intelligent Dependable Embedded Architectures Lab University of Kentucky
Runtime Behavior • Local CAN Bus Network Interface management tasks: – Scheduler Schedule – Network (high priority) Interface Memory Loader (low priority) – Module Loader Altitude “inputs” • Module I/O data (Q2) Elevator passed through Control Low Qual. state info Altitude Est. mailboxes. Messages • Data routing is to System Manager Next Release elevator cmd transparent to angle Modules. Servo 1 driver Chute logic Chute release 10/11 Intelligent Dependable Embedded Architectures Lab University of Kentucky
Current Research • Expand bus via wireless link to the Sensors ground: Actuator Autopilot Interface – Rapid prototyping Mission Resource Control Manager – Minimize risk to hardware Mission Payload Interface – Flexible Reconfiguration Wireless Mission Link Payload I/O • Applying the Wireless Link framework to the Experimental Autopilot design of BIG BLUE IV `1 Flight Data Analysis Pilot-in-the-loop Aircraft 11/11 Intelligent Dependable Embedded Architectures Lab Simulator University of Kentucky
Recommend
More recommend
