Projects


Active Projects


Android apps for mobile robots for autonomous robotic tasks
Vivek Yogi, Arpita Sinha, Leena Vachhani

Modern Smartphones are equipped with numerous sensors, multiple communication devices, user interfacing mechanisms and a polished operating system. By adding a minimal hardware infrastructure for actuation we can create a standalone robot agent. The external hardware infrastructure required can be interfaced to the Smartphone using one of the standard communication interfaces. Such a design offers flexibility of actuation design for implementing different applications. Such a design also offers scope for scaling of the robot infrastructure for future needs. I work on enabling practical applications with such a robot. Design and implementation of real-time capable algorithm forms the core part of my work. The implemented applications are in the form of Apps that can be easily downloaded and put to use by anyone with an Android smartphone.


Target tracking with range-only information
Gaurav Chaudhary, Twinkle Tripathy, Aseem Borkar, Arpita Sinha

Here we implement some control strategies on differential drive robots for tracking both stationary and moving targets, using only range information, i.e., bearing information to the target is not available. Some candidate controllers including continuous and switching strategies have been implemented to achieve this objective. It is also ensured that these controllers adhere to a set of theoretical conditions for capture of stationary and moving targets. In the experiments in the adjacent video the Firebird V robot acts as the pursuer and the spark V robot acts as the target. The localisation of the robots is done using a MATLAB based visual feedback system which uses four calibrated web cameras to detect a unique colored pattern mounted on the robots to localize them in the experimental area.


Collision-free multi-agent formation on Lissajous curves for area surveillance
Aseem Borkar

A multi-agent formation is proposed that moves on a Lissajous curve in a collision-free manner. This multi-agent formation simultaneously performs multiple surveillance tasks such as repeated collision-free area surveillance, guaranteed complete sensor coverage of the rectangular area, and finite time target detection and entrapping. The agents are considered have finite non-zero dimensions and a sufficient upper bound on the agent dimensions is derived to guarantee collision free motion of the formation on the Lissajous curve. We have also proposed an algorithm to select the number of agents to be deployed, and the optimal Lissajous curve to be used (in terms of ares coverage time and bound on agent size). This strategy is easily scalable and doesn't need any cooperation among the agents as collision-free motion is achieved simply by leveraging the properties of the Lissajous curve. The validation of this strategy both in MATLAB simulation and experiments can be found in the video.


Reconfigurable quadrotor formations on Lissajous curves for area surveillance
Aseem Borkar

The previous idea of a non-cooperating multi-agent formation on a Lissajous curve is extended to a cooperating and reconfigurable multi-quadrotor formation on Lissajous curves for collision-free aerial surveillance of a rectangular area. The reconfiguration strategies are proposed for quadrotor addition, removal or replaced using a decentralized cooperating scheme. We have validated our results through MATLAB simulations for agents having a non-zero size satisfying a theoretically derived size bound. To demonstrate the practical applicability of the proposed surveillance and reconfiguration strategies, we have performed simulations for quadrotors in a ROS-Gazebo based Software-In-The-Loop (SITL) simulator and have implemented the same with a team of five indigenously developed quadrotors operating in a Vicon motion capture environment. The videos of the SITL simulation and the experiment can be found below.


Completed Projects


Pattern switching between annular regions for a unicycle agent
Twinkle Tripathy, Aseem Borkar, Arpita Sinha, Hemendra Arya

Here we implement a control law proposed for generating annular patterns centered about a stationary target point, capable of switching between different annular regions in run time for a agent modeled using unicycle kinematics. The control law is implemented using a Firebird V differential drive robot. The robot is localized in the experimentation area by a visual feedback system (developed in house) using a calibrated wide-angle lens camera. The localisation information with time stamp is communicated to the robot using XBee modules. The annular region bounds in the video turn green to indicate switching and the active bounds are shown in red.

 Consensus based Deviated Cyclic Pursuit for Target Tracking Applications
 A Study of Balanced Circular Formation under Deviated Cyclic Pursuit Strategy


Relative heading based pattern generation
Shashank Agarwal, Twinkle Tripathy, Aseem Borkar, Arpita Sinha

Here a unicycle based kinematics for the agent generates patterns with respect to a fixed target point,using a simple non-linear relative heading based control law. The control law is implemented using a Firebird V differential drive robot. The robot is localized in the experimentation area by a visual feedback system (developed in house) using a calibrated wide-angle lens camera. The localisation information with time stamp is communicated to the robot using XBee modules. The cases considered for experiments shown in the adjacent video densely scan the annular region bounding the patterns generated. Thus the video only shows parts of the experiments.


Aerial convoy monitoring using elliptical orbits
Aseem Borkar, Swaroop Hangal, Hemendra Arya, Arpita Sinha, Leena Vachhani

A novel vector field based guidance scheme is proposed that makes an aerial agent monitor a convoy of targets moving along a possibly non-linear trajectory on the ground. The scheme first computes a time varying ellipse that encompasses all the targets in the convoy using a simple regression based algorithm. It then ensures convergence of the aerial agent to a trajectory that repeatedly traverses this moving ellipse. The correct tracking behaviour of the scheme is rigorously established along with error bounds, based on the perturbation theory of ordinary differential equations. It is supported by MATLAB simulations and experiments using the AR.Drone 2.0 as the aerial agent and Firebird V robots as the ground convoy in a Vicon motion capture environment. The videos of the simulations and experiments are shown in the videos.


2D Localization system based on visual feedback from cameras
Aseem Borkar

This a simple visual feedback system developed using MATLAB for localization ground robots in a 2D arena for both lab courses and research experiments. The initial version comprised of a single web cam for localisation of robots in a 1 meter X 1 meter square area. The robots were equipped with unique colored patterns for easy identification. This system was later expanded to a four camera system for greater area. The four cameras were later replaced by a single wide-angle lens camera and instead of color patterns unique black and white identification patterns were used for identifying robots in the image processing code. The localisation data including the (x,y) position coordinates and the heading angle are transmitted to the robots using wireless Xbee network operating in the API mode that supports addressed communication packets. A block diagram of the system is shown below along with a video showing its different stages of development.


Experiments designed for Control Lab course
Aseem Borkar

These are some of the experiments designed for the SC 626 Controls Lab course (Spring 2017), using the AR.Drone 2.0. The drone is controlled using the ardrone_autonomy package in ROS. Some simple control task implemented were: basic way-point navigation with fixed heading, way-point navigation treating the drone as a unicycle, following a virtual leader point moving on a circular orbit, elliptical orbit tracking using vector field guidance and tracking a moving ground target.


ROS enabled autonomous quadrotor for experiments in Vicon motion capture environment
Aseem Borkar, Swaroop Hangal, Vraj Parikh, Shoeb Ahmed, Hemendra Arya

We built ROS enabled quadrotor for indoor multi agent experiments in a Vicon motion capture environment. The quadrotor has a flight time of up to 14 minutes and is equipped with a Pixhawk v1 flight controller and the Raspberry Pi 3B as a companion computer which runs the ROS-Node to control the drone. The localisation data from the Vicon system is communicated to the drone using a wi-fi network. A labeled photo of the quadrotor with its components is given below and the various stages of its development and testing are shown in the adjacent video.


ROS-Node for driving multiple Firebird V robots
Aseem Borkar, Vraj Parikh, Arpita Sinha

We made a python based ROS-Node to control multiple Firebird V robots in a Vicon motion capture environment. The ROS-Node takes number of robots as an input argument and opens command velocity topics in ROS for each robot. The velocity command published to these topics are then transmitted to the corresponding robots using Xbee Pro series 1 wireless modules (also mounted on the robots). Each robot Xbee is configured with a unique address, and the API packet format supported by the XBee firmware is used to communicate addressed packets to each robot.


Swarm Aggregation without communication and global positioning
Dhruv Shah, Leena Vachhani

In this work, we propose a novel decentralized controller for the aggregation of a swarm of unicycle agents in the absence of communication and exchange of global positions. The methodology implicitly facilitates tasks like obstacle avoidance and path-following, while the agents exercise local decision-making. The primary aim of this work is to form stable aggregates of a swarm using only local sensing and no inter-agent communication. A theoretical analysis of the proposed algorithm is presented based on the theory of switched systems, establishing asymptotic stability of the formed aggregates. Simulation experiments are presented to show the stability of aggregates in the presence of large, extended obstacles and external disturbances. Furthermore, we demonstrate an implementation of the proposed controller on a real system using experiments on a swarm of differential drive microbots. The proposed technique without communication demonstrates similar performance when compared with existing swarm aggregation techniques using global and local communications.


Trochoid pattern formation using multiple robots
Jerome Moses, Arpita Sinha

Patterns have been vividly displayed in nature in different forms. One such pattern is trochoids. Suppose we want mobile robots to trace these patterns. What kind of control laws will ensure that the trajectories of the robots resemble these patterns? In our work, we design control law for multiple robots tracing the trochoidal patterns in synchronization with each other. Here the robots do not follow a pre-specified trajectory. Rather the trajectories of the robots evolve through the interaction of the robots using linear consensus protocol. These patterns are not only aesthetically appealing but can also be used in applications such as like search, exploration or surveillance.


Path planning for minimal energy usage
Manas Choudhari, Leena Vachhani

A new method for energy efficient trajectory planning using Dubins path having linear time complexity with respect to the number of waypoints has been proposed. Although it does guarantee positive energy savings, the average energy savings are 14.8% compared to an existing trajectory planner which uses B´ezier curves. Conventional waypoints planning algorithms such as A* fail to generate optimum results when used with cost functions involving turns. An Edge based search approach has been proposed which considers edges as the nodes during search. Edge based A* and Edge based Theta* methods which are the enhanced versions of A* and Theta* algorithms have been proposed. The proposed Edge based A* and Edge based Theta* methods provide average energy savings of 20.8% and 29.7% compared to their conventional counterparts. The comparison of different methods proposed has shown that waypoints planning has a greater impact on energy savings compared to trajectory planning.


Contour Formation using Boundary Tracking with Mobile Robots
Prahlad Das, Arpita Sinha

In present days, many human operations are being replaced by robots. These robots are efficient, less expensive and work in lesser time than human. But using a single robot to a specific task might be less efficient and prone to failure. That’s why nowadays people use multi-robots rather than single. Boundary tracking is nowadays an active field of research. Because of its application in many areas researchers are looking for new methods for boundary tracking and following. Using multiple robots for boundary tracking can solve a lot of problems in disaster management and protection of people. In this report, we purpose an algorithm to find out contours of a source, then track these contours using multiple mobile robots such that they do not collide as well as do not go out of communication range. All these robots should change the contours to move nearer to source. MATLAB simulation is done for the algorithm. Further algorithm is implemented on Kilobots.


Monitoring a target with multiple autonomous vehicles
Sangeeta Daingade, Arpita Sinha

This work addresses a problem of monitoring a target with multiple autonomous vehicles where the objective is to make the vehicles move around a target with uniform distribution. An uniform distribution of vehicles along a circular orbit is usually desirable in order to provide the best overall coverage of the target and its surroundings. It is also important to develop simple control strategies that can be easily implemented on actual vehicles. This work presents decentralized control strategies for unicycle agents that are simple, easy to compute and require easily available information.


Multi-Robot Graph exploration and Mapping
Sarat Chandra Nagavarapu, Leena Vachhani, Arpita Sinha

In this work, we have proposed a decentralized multi-robot coordinated exploration and mapping strategy, called Improved Multi-Robot Depth First Search (IMRDFS). A modified incidence matrix (MIM) data structure is introduced to store and merge the independently generated maps. A new map merging technique to integrate the information of two or more local maps stored in the form of MIMs is also proposed. Besides building the map of the environment, this data structure (MIM) also facilitates coordination among the robots, minimizes exploration redundancy, accelerates exploration completion and declaration of task completion. A generic framework using the MIM data structure is proposed for any initial placement of the agents, search strategy and information exchange method. A generic approach is developed to obtain conditions for the finite time completion of the exploration task. Multiple robots maneuvering in an environment must avoid collision with other objects in the environment as well with each other in order to guarantee an efficient task completion. Collision between robots while exploring a graph occurs when the robots follow a common edge or reach to a common vertex. In order to avoid a collision, the robots must have the capability to detect a collision. The probability of collisions increases with the increase in the number of robots operating in the environment. Obtaining a collision condition in the absence of inter robot communication using minimal sensing is a major challenge in collision avoidance. In this work, we have proposed a speed variation strategy to avoid inter-robot collisions. It is shown that minimal sensing is sufficient to detect the possibility of collision in a decentralized system. A necessary and sufficient condition for collision is found. When a possible collision is detected, the proposed strategy of changing speeds of the robots by predefined values is proved to avoid collision.



Coordinated Nodding of a 2D Lidar for Dense 3D Range Measurements
Anindya Harchowdhury, Leena Vachhani

Sensing technology is an integral part of any autonomous system. In particular, 2D range sensors have played a critical role in localization, mapping and other sensing tasks in robotics. But, 2D measurements provide limited information about the environment, while 3D range sensors provide a better understanding of the environment for the robot to make crucial decisions. However, 3D range sensors presently available in market are too costly to be deployed by small robots and in large numbers. Therefore, we introduce a novel low cost 3D range sensor, developed using an affordable 2D lidar. A high speed servo is used to nod the lidar and generate 3D perception of surrounding environment. We developed a mathematical model of the dynamics of the sensor system to prove results of unique point set registration of a 3D surface while mounted on a robot. The proposed 3D laser sensor is novel in the way the horizontal and vertical angle are coordinated to produce unique dense scans. In addition, we propose an optimal nodding scheme such that the laser beams cover the whole scan window in the best possible manner. Real time environment reconstruction has been performed by mounting it on a mobile robot within an indoor environment.

 Coordinated Nodding of a 2D Lidar for Dense 3D Range Measurements



Interval Analysis methods for Collision detection and avoidance
Pranjal Vyas, Leena Vachhani

Collision detection and avoidance are vital sub-tasks in any autonomous robotic task. In this work, we develop a technique that guarantees collision avoidance in an environment containing multiple agents. These agents can be of different types such as static, dynamic obstacles and other robots. The technique supports parallel implementation and is appropriate for real-time applications. We use the theory of interval arithmetic to represent the pose of agents as intervals in a fixed time period. Geometrically, the intervals correspond to finite-length arcs and line-segments. Theoretical results on the inclusion of one interval, in another interval, in terms of subintervals are derived. Breaking down the problem in sub-intervals supports parallelism in performing multiple interval inclusion tests and in handling multiple agents. The proposed interval-arithmetic based framework leads directly to a hardware-efficient collision detection scheme. In particular, the proposed strategy admits a solution even for a dynamic environment using just shift and add capability, an important aspect for embedded implementation. The solution of interval inclusion is also used to find a set of solutions for guaranteed collision avoidance with multiple agents with known or unknown trajectories. Simulation results in MATLAB and experiments with an FPGA driven differential drive mobile robot demonstrate the versatility of the proposed approach.



Interval Analysis methods for Optimal Path Planning
Pranjal Vyas, Leena Vachhani

Path planning of autonomous mobile robots in cluttered environment is a challenging and a well studied problem. The configuration space based methods are typically used for finding out the feasible and non-feasible areas for a mobile robot. Finding an optimal path considering admissible control inputs is challenging as it needs to explore all the possible options, present and future. The interval analysis approach can used for dividing the configuration space into feasible and non-feasible spaces. This technique is capable of dividing the space into intervals of different sizes which helps to cover the obstacles in a refined manner. This paper presents the computation of collision-free path using interval analysis between any two points in a 2D-environment cluttered with obstacles for a non-holonomic robot. The admissible control inputs for the robot are considered as constraints and are included in the configuration space parameters while computing the collision-free regions. The optimal shortest path between two points is computed using A* search algorithm and Pontryagin's minimum principle. Simulation results are shown using MATLAB by computing the path between two points considering selected control inputs as constraints.