Hello, I am Akash
I am an enthusiast continuously working towards integrated intelligent systems. The key focus areas include robot system design and integration with sensors like Camera, Radar, GPS etc .
Behaviour based Control Strategy for Double Ackermann Steering Control of Autonomous Tandem Road Rollers
Objective
The project work is an attempt to explore the autonomous capabilities of tandem rollers in the road construction. The project specifically focuses on analyzing the dynamics and kinematics of tandem rollers, understanding the previously developed work on autonomous navigation including different levels of controls based on perception; later, extending the same to the edge compaction module of the navigation system. The manual edge compaction for a roller is a complicated task which requires a skilled driver to drive through the edge with the front edge cutter and keeping the rear wheels away from the edge. The autonomous edge controller is an attempt to remove the driver involvement with the automated crab walk application in the roller.
Workflow
- Analysis of Behaviour based Control Strategy
- Study of control architecture and communication protocol
- Road representation
- Edge extraction
- Trajectory creation
- Crab angle calculation
- Trajectory tracking with crab behaviour activated
- Edge Cutter activation
- Simulation using Unreal Engine and Finroc (ROS inspired tool)
Backgroud
Integrated Behaviour based Control Strategy (iB2C)
[Proetzsch 10] proposed a behaviour based control architecture iB2C for the development of complex robotic systems. “It is shown how architectural principles support several behaviour based mechanisms, e. g. coordination mechanisms, behaviour interaction, and hierarchical abstraction”.
- Structure and Components
- Behaviour Fusion
- - Maximum Fusion
- - Weighted Average Fusion
The iB2C behaviour module (B) consists of six signals coming in and out namely, Stimulation s , Inhibition vector i , Input vector e⃗, Output vector u⃗, Activation vector a⃗ and Target rating r . The stimulation s is responsible for a gradual initiation of behaviour, while the inhibition vector i⃗ gradually disables it. The combined interaction of s [0, 1] and i⃗ [0, 1] triggers an activation l signal which sets the limiting value for the influence of behaviour.
l = s (1 - i )
The activation signal ( l ), further helps in defining a transfer function F ( e⃗, l ) which later defines the respective output vector u⃗ of the system i.e. F ( e⃗, l ) = u⃗. The data from sensors, (e.g. distance measurements, acceleration) also forms an input vector e ∈ Rm . The activation l also limits the activity a of the behaviour B, such that l≤ a . Behaviour B can thus be described as:
B = ( fa , fr , F )
where, fa is the activity function, fr is the target rating function, F is the transfer function of the behaviour. It is interesting to note that, like the transfer function, activation function fa is a function of e⃗ and l ; and also defines activity signal a and activity vector a⃗ .
The basic design of iB2C Behavior and Percept module
Behaviour group containing iB2C modules and iB2C fusion modules
Behaviour fusion can be achieved using following techniques:
Control Architecture Integrating Different Level Of Navigation Tasks
REACTiON control architecture with Remote Interface [Wolf 18][Ropertz 18b]
Remote Interface based on 5G AMMCOA (Autonomous Mobile Machine Communication for Off-Road Applications) Protocol
Related Work
Road Representation and Edge Extraction
Road R representation with Seam S⃗ , Lane L⃗, Edge E⃗, Grid G⃗ and Tracks T⃗ [Wolf 19]
Bicycle Model
Concept and Implementation
Schematic representation of the approach
Edge Compaction Behaviour
Trajectory Decision based on Behaviour Fusion
Trajectory Tracking
- Pilot
Pilot decides the maneuvering parameters
- Steering Controller
Receives the maneuvering parameters from the Pilot and transforms them into wheel angles and velocity
Controller input to ackermann steering hardware
Results
Simulation on Unreal Engine
Simulation on FinGUi
