Implementation of HumanLike Driving Skills by Autonomous Fuzzy
Implementation of Human-Like Driving Skills by Autonomous Fuzzy Behavior Control on an FPGA-Based Car-Like Mobile Robot by Tzuu-Hseng S. Li, Shih-Jie Chang and Yi-Xiang Chen An Article Review by Matthew D. Schrieber
Outline • Article Objectives • Overview of Different Controllers • Simulated Results • Circuit Design on an FPGA • Real-Time Experiment Results • Contributions • Comments and Criticisms 2
What is a Car-Like Mobile Robot (CLMR)? • The rear wheels are fixed parallel to the car body • The front wheels can turn to the left or right, but must remain parallel to each other 3
What are Human-Like Driving Skills? 4
What is Fuzzy Logic? • Computer logic based on Degree of Membership (DOM) ranging from 0 to 1 instead of true or false (0 or 1) • Membership functions are used to describe variables. • Uses linguistic variables to describe the partitions of the membership function Retrieved from: http: //en. wikipedia. org/wiki/Fuzzy_control_system 5
What is a Fuzzy Logic Controller (FLC)? • Preforms that same function of a traditional controller Retrieved from: http: //en. wikipedia. org/wiki/Fuzzy_control_system • Inputs require fuzzification and outputs require defuzzification to become crisp again • Uses specific rule sets to determine output eg. IF temperature IS cold THEN heater IS on 6
Article Objectives • To merge the concepts of car maneuvers, fuzzy logic controls, and sensor-based behaviours • Implement an autonomous car-like mobile robot capable of human-like driving skills 7
Autonomous Fuzzy Behaviour Controller The CLMR’s behaviour is broken down into the following 4 different fuzzy controllers: • Fuzzy wall-following control (FWFC) • Fuzzy corner control (FCC) • Fuzzy garage-parking control (FGPC) • Fuzzy parallel-parking control (FPPC) Notes: • The FWFC is the only controller that is entirely fuzzy based • The other controllers have addition algorithms that use the FWFC. 8
Fuzzy Wall-Following Control 9
Fuzzy Wall-Following Control Steering Angle • Forward Motion • Backward Motion Speed Control 10
Fuzzy Wall-Following Control Membership Functions • Steering Angle Control • Speed Control 11
Fuzzy Wall-Following Control Fuzzy Rule Tables Steering Angle Control Speed Control 12
Fuzzy Corner Control 13
Fuzzy Garage-Parking Control 14
Fuzzy Parallel-Parking Control 15
Behaviour Selection Mechanism The wall-following mode is the default behaviour, however the behaviour will switch depending on the measured values of the sensors. The selection rules are as follows: > > >> >> 16
Simulated Results Notes: • It was found that the CLMR could not correctly drive into the parking space if the membership function for the steering angles were partitioned into 3 instead of 5. • There was no obvious different between 5 and 7 partitions. 17
Hardware Architecture FPGA: Altera Flex series manufactured by Galaxy Far East CAD Tool: MAXPLUS II Sensors: UF 66 MG manufactured by TELCO International A/D Converter, D/A Converter and Motor Driver IC. 18
Circuit Design on an FPGA Inputs pins: 8 for each sensor via A/D converter and 1 for the clock Output pins: 2 for DC motor driver, 1 for A/D converter, 1 for DC servomotor, 8 for the D/A converter 19
Behaviour Selection Control Module • Uses a Mux to implement the Behaviour Selection Mechanism described earlier • Both outputs describe whether the CLMR is going forwards or backwards 20
Pulse-Width Modulation Fitness Module • Used for controlling the DC servomotor that controls the steering angle. • Using the Crisp output from the FLC 21
Design of FLC Module • Steering angle and speed control both use this module 22
Fuzzification Submodule • Sensor-based inputs • “Choice” signal indicates whether CLMR should go forward or backward 23
Decision-Making Logic Submodule • The inference rule base block is realized in a lookup table 24
Defuzzification Submodule • Done by using the Weighted Average Method • Parallel multiplier used to speed up performance 25
Real–Time Experimental Results Capable of autonomously maneuvering in their test ground which includes garage and parallel parking CLMR Dimensions: • Length 380 mm • Width 240 mm • Weight 5 kg 26
Contributions • Great foundation work for applying Fuzzy Logic Controller to mobile robotics • Autonomous behavior controllers capable of human-like driving skills are becoming features in modern cars (Automatic Parallel Parking) 27
General Comments and Criticisms • Proficiency in English was lacking • Many decisions lacked justification • Significant details about the development of many of the individual components was lacking • Impressive merger of many different design concepts 28
Specific Comments and Criticisms • The functionality of the Behaviour Selection Mechanism is unclear. • Other portions of the different controllers could have been implemented using fuzzy logic instead of just the FWFC. • The simulated and experimental results are impressive and the varying of the membership function partitions was insightful. 29
Questions? 30
- Slides: 30