Interested students:  Please contact me directly. I’m also open to discuss project suggestions of your own.

Part III / MPhil ACS Project Proposals

  1. The Blind Robot Racecar
  2. Private Formation Control for Teams of Mobile Robots
  3. Ensuring Passenger Privacy in Mobility-on-Demand Systems
  4. Optimal Vehicle Assignment Algorithms for Heterogeneous Fleets in Mobility-on-Demand Systems
  5. Hybrid Assignment Algorithms for the Distribution of Heterogeneous Mobile Robots
  6. Formalization of Heterogeneity in Multi-Robot Systems

Part II Project Proposals

  1. The Robot Race
  2. Leader-Follower Robot Formations


The Blind Robot Racecar.

To overcome autonomy challenges, the standard approach in mobile robotics research is to over- provision the mobile platforms with sophisticated on-board sensors, including laser-range finders, cameras, and radars. The potential of off-board collaborative sensing, through sensors embedded in static infrastructure, remains largely unexplored. This project poses the research question the other way around: How well can a robot perform when it owns nothing but a communication module, and relies fully on off-board sensing to navigate a path and avoid collisions? Consequently, the goal of this project is to empower a ‘blind’ race car to navigate from start to goal (as quickly as possible) by instrumenting the race course with a fixed sensor network. The study aims at finding the minimal amount of sensors necessary to make this possible. This will involve exploring system design, distributed estimation and navigation algorithms. Initially, the project will be developed in the realistic robot simulator Gazebo [1]. If time allows, real-robot experiments will be conducted to validate the findings.

References:
[1] http://gazebosim.org/
[
2] http://www.ros.org/

Private Formation Control for Teams of Mobile Robots.
Navigation in formation lies at the core of most multi-robot applications, including vehicle platooning, mobile surveillance, or warehouse automation. Indeed, formation control algorithms allow the robots to maintain local communication networks, to collaboratively handle objects, or to improve the team’s navigation efficiency (e.g., through highway platooning). A common approach toward formation control consists of selecting one or more leading vehicles to direct the trajectory [1]. However, doing so poses a security threat: if the leading vehicles are compromised, the whole robot formation can be lead astray. The idea of this project is to explore privacy techniques that protect the roles of the various robots in formation [2]. Specifically, the goal of this project is to develop multi-robot navigation algorithms that ‘hide’ the leading vehicles, and obfuscate the underlying formation control mechanisms. The findings are to be validated in the realistic robot simulator Gazebo [3]. If time allows, real-robot experiments will be conducted to demonstrate the efficacy of the developed approach.

References:
[1] J. P. Desai, J. Ostrowski, V. Kumar; Controlling Formations of Multiple Mobile Robots; IEEE International Conference on Robotics and Automation (ICRA), pp. 2864–2869, 1998.
[2] A. Prorok, V. Kumar, A Macroscopic Privacy Model for Heterogeneous Robot Swarms, International Conference on Swarm Intelligence, 2016.
[3] http://gazebosim.org/
[4] http://www.ros.org/

Ensuring Passenger Privacy in Mobility-on-Demand Systems.
Urban transportation is being transformed by mobility-on-demand (MoD) systems. One of the goals of MoD systems is to provide personalized transportation services to passengers (e.g., Uber, Lyft). This process is facilitated by a centralized operator that coordinates the assignment of vehicles to individual passengers, based on location data. In recent work, by obfuscating vehicle positions, we propose a linear sum assignment algorithm that ensures passengers’ location privacy [1]. Our results demonstrate the feasibility of integrating privacy into shared mobility systems. Yet, many questions remain to be explored. In particular, the goals of this project are to explore ways of providing individualized privacy, as a function of heterogeneous user preferences and habits.

References:
[1] A. Prorok, V. Kumar, Privacy-Preserving Vehicle Assignment for Mobility-on-Demand Systems, IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017; https://arxiv.org/abs/1703.04738

Optimal Vehicle Assignment Algorithms for Heterogeneous Fleets in Mobility-on-Demand Systems.
Urban transportation is being transformed by mobility-on-demand (MoD) systems that consider fleets of vehicles and passengers demanding to be picked up at specific locations. Formally, this problem can be posed as a batch assignment of vehicles to passengers such that passenger waiting times and travel distances are minimized [1, 2]. However, prior work considers homogeneous vehicle fleets. The goal of this project is to develop optimal allocation algorithms that can deal with vehicles that have heterogeneous and potentially unknown traits, such as passenger capacities and traveling speeds. Results are to be validated on relevant datasets, such as the Manhattan Taxi Dataset [3].

References:
[1] A. Prorok, V. Kumar, Privacy-Preserving Vehicle Assignment for Mobility-on-Demand Systems, IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017; https://arxiv.org/abs/1703.04738
[2] J. Alonso-Mora, S. Samaranayake, A. Wallar, E. Frazzoli, and D. Rus. On-demand high-capacity ride- sharing via dynamic trip-vehicle assignment. Proceedings of the National Academy of Sciences, 2017. doi: 10.1073/pnas.1611675114.
[3] NYC Taxi & Limousine Commission, Trip Record Data, http://www.nyc.gov/html/tlc/html/about/trip_record_data.shtml

Hybrid Assignment Algorithms for the Distribution of Heterogeneous Mobile Robots.
Multi-robot task allocation is a general approach to coordinating a team of robots that need to complete a set of tasks collectively. Solutions to this problem find a number of real-world applications, such as mobility-on-demand systems, and mobile surveillance. Yet, when the set of robots is heterogeneous and when individual tasks require multiple robot types simultaneously, the assignment problem is NP- hard [1]. By taking a mean-field approach to the problem, more efficient strategies can be found. However, these strategies lack the precision of discrete assignment algorithms. The goal of this project is to explore hybrid methods that are computationally efficient (polynomial in the number of robots and tasks), and that approximate solutions provided by discrete solvers.

References:
[1] B. P. Gerkey and M. J. Mataric. A Formal Analysis and Taxonomy of Task Allocation in Multi-Robot Systems. Interntional Journal of Robotics Research, 23(9):939– 954, 2004.
[2] A. Prorok, M. A. Hsieh, and V. Kumar. The Impact of Diversity on Optimal Control Policies for Heterogeneous Robot Swarms. IEEE Transactions on Robotics (T-RO), vol. 33, no. 2, pp. 346 – 358, DOI: 10.1109/TRO.2016.2631593. April 2017

Formalization of Heterogeneity in Multi-Robot Systems.
Heterogeneous robot teams an provide advantages over homogeneous robot teams. Indeed, heterogeneous robots leverage complementary capabilities to collaborate and solve more complex tasks. However, as heterogeneity becomes a design feature, we need new analytical tools that enable us to build such systems optimally. In recent work, we were the first to formalize heterogeneity for robot systems [1]. Yet, this representation lacks in versatility and realism. The goal of this project is to extend the formalism to capture essential robotic characteristics, such as uncertainty, continuity, and time- variability.

References:
[1] A. Prorok, M. A. Hsieh, and V. Kumar. The Impact of Diversity on Optimal Control Policies for Heterogeneous Robot Swarms. IEEE Transactions on Robotics (T-RO), vol. 33, no. 2, pp. 346 – 358, DOI: 10.1109/TRO.2016.2631593. April 2017

The Robot Race.
The goal of this project is to develop basic algorithms so that a mobile robot (e.g., Turtlebot [1]) can successfully navigate from the start of a race course to its goal. In order to do this, we will leverage available algorithm repositories, such as [2], that provide basic perception and navigation functions. If multiple students sign up for this project, an end-of-project race will be organized.

References:
[1] http://www.turtlebot.com/
[2] https://github.com/mit-racecar
[3] http://www.ros.org/

Leader-Follower Robot Formations.
Navigation in formation lies at the core of most multi-robot applications, including vehicle platooning, mobile surveillance, or warehouse automation. Indeed, formation control algorithms allow the robots to maintain local communication networks, to collaboratively handle objects, or to improve the team’s navigation efficiency (e.g., through highway platooning). A common approach toward formation control consists of selecting one or more leading vehicles to direct the trajectory [1]. The goal of this project is to deploy leader-follower algorithms on teams of real robots (e.g., Turtlebots [2]), and to integrate a user interface (such as a joystick), allowing users to direct the trajectory of the robot formation.

References:
[1] J. P. Desai, J. Ostrowski, V. Kumar; Controlling Formations of Multiple Mobile Robots; IEEE International Conference on Robotics and Automation (ICRA), pp. 2864–2869, 1998.
[2] http://www.turtlebot.com/
[3] http://www.ros.org/