DAY 2: Ph.D Forum

Attacks on a set of agents with cooperative and antagonistic interactions attempting to achieve linear bipartite consensus is considered here. In bipartite consensus, all the agents converge to a final state characterized by identical modulus but opposite sign. Adversary seeks to slow down the bipartite consensus by a Denial of Service (DoS) type attack where the attacker has the capability to break a specific number of links at each time instant. The problem is formulated as an optimal control problem and the optimal strategy for the adversary is determined. This is illustrated by a numerical simulation of a four agent system.

UAV (Unmanned Aerial Vehicle) networks, due to their versatility and applicability have attracted significant attention in the past decade. UAVs carry a wide range of sensors with it. The information it collects is either communicated directly to ground station or synced with cloud. Conflict resolution is an indispensable aspect of multi-UAV airspace achieved by processing the sensor data to compute evasive actions. This paper presents conflict resolution using correlated equilibrium which is a game theoretic decision-making module. The framework defines a semi-autonomous system capable of handling multi-UAV networks. The data perceived by the UAVs are communicated to the ground station for conflict resolution. This integrates the cyber processes into physical devices. The CONCORD framework could be integrated with any available UAV traffic management system for fair and guaranteed conflict resolution.

An aerial robot’s adaptive feedback linearization control design in a cyber-physical system is studied in this work. The aerial robot consists of a quadcopter attached to a three-degree of freedom manipulator. It forms a quadcopter manipulator system (QMS). The autopilot in the QMS uses on-board sensor data, computes control inputs based on the proposed control methodology, and makes decisions on actuating the quadcopter manipulator system. The network of QMS is given an aerial painting task in this project. The task divides the entire painting area into small squares, with the center designated as the QMS’s goal location. The coupling between the two subsystems, the quadcopter and robotic arm, creates counter torques and moments on each other while executing the job. In addition, because of the unknown environment, the task assigned to each QMS may change during the operation. The proposed work develops an adaptive augmented torque control law to enable system stability and trajectory tracking control in the provided scenario. The update law for the estimation of unknown system parameters is developed using the strictly positive real-Lyapunov method. The proposed controller is then implemented and validated in ROS/Gazebo and MATLAB simulators.

Collaborative load transportation by autonomous teams of unmanned aerial vehicles (UAVs) is an emerging application of cyber-physical systems. We address a problem where a team of UAVs transports a semi-flexible payload in an environment with obstacles. The consideration of a semi-flexible payload allows for flexibility in safe-shape manipulation and navigation between obstacles. Control barrier functions (CBFs) are used to construct the obstacle avoidance and shape constraints in an optimization-based control problem. The analysis is supported with simulation results for a team of four UAVs manipulating a payload in the presence of obstacles and settling on a standoff circle about a target.

Reinforcement learning (RL) in real world applications faces major hurdles – the foremost being safety of the physical system controlled by the learning agent and the varying environment conditions in which the autonomous agent functions. A RL agent learns to control a system by exploring available actions. In some operating states, when the RL agent exercises an exploratory action, the system may enter unsafe operation, which can lead to safety hazards both for the system as well as for humans supervising the system. Additionally, RL autonomous agents learn optimal decisions in the presence of a stationary environment. However, the stationary assumption on the environment is very restrictive. In many real world problems like traffic signal control, robotic applications, etc., one often encounters situations with non-stationary environments, and in these scenarios, RL algorithms yield sub-optimal decisions. In my doctoral thesis, I develop algorithms that tackle these challenges to real world application of RL algorithms. My thesis also develops and solves RL models for a resource sharing problem in industrial internet of things (IIoT) and the obstacle avoidance problem in unmanned aerial vehicle navigation.

We propose a novel deep learning framework that focuses on decomposing the motion or the flow of the pixels from the background for an improved and longer prediction of video sequences. We propose to generate multi-timestep pixel level prediction using a framework that is trained to learn the temporal and spatial dependencies encoded in video data separately. The proposed framework called Velocity Acceleration Network or VANet is capable of predicting long term video frames for the static scenario, where the camera is stationary, as well as the dynamic partially observable cases, where the camera is mounted on a moving platform (cars or robots). This framework decomposes the flow of the image sequences into velocity and acceleration maps and learns the temporal transformations using a convolutional LSTM network. Our detailed empirical study on three different datasets (BAIR, KTH and KITTI) shows that conditioning recurrent networks like LSTMs with higher order optical flow maps results in improved inference capabilities for videos.