Private Distributed Optimization

In a distributed machine learning scenario, the dataset is stored among several machines and one needs to solve a distributed optimization problem to collectively learn the underlying model. We present a privacy preserving distributed algorithm for optimizing a convex function consisting of several possibly non-convex functions. Each individual objective function is privately stored with an agent while the agents communicate model parameters with neighbor machines connected in a network. We show that our algorithm can correctly optimize the overall objective function and learn the underlying model accurately. We further prove that under a vertex connectivity condition on the topology, our algorithm preserves privacy of individual objective functions. We establish limits on the what a node can learn by observing the messages and states shared over a network.

  • S. Gade, N. H. Vaidya. “Privacy-Preserving Distributed Learning via Obfuscated Stochastic Gradients”, IEEE Conference on Decision and Control, Miami Beach, FL, 17-19 December, 2018.  (pdf)
  • S. Gade, N. H. Vaidya. “Private Optimization on Networks”, IEEE American Control Conference, Milwaukee, WI, 26-29 June, 2018. (pdf) (link)
  • S. Gade, N. H. Vaidya. “Private Learning on Networks: Part II”, arXiv:1703:09185. (pdf)
  • S. Gade, N. H. Vaidya. “Private Learning on Networks”, arXiv:1612.05236. (pdf)
  • S. Gade, N. H. Vaidya. “Distributed Optimization of Convex Sum of Non-Convex Functions”,  arXiv:1608.05401. (pdf)
  • S. Gade, N. H. Vaidya. “Distributed Optimization for Client-Server Architecture with Negative Gradient Weights”, arXiv:1608.03866. (pdf)

Robotic Herding 

In this work we develop an active technique for herding of a group of animals/robots using robots. This technique is meant for controlling and tackling organized bird activity in the form of flocks. We formulate a non-linear coupled flocking algorithm based on nearest neighbor interaction and augment it with evasion laws inspired from empirical predator-prey behavior. A novel herding technique, called the n-wavefront herding algorithm is developed. In this boundary control type strategy, the pursuer influences only the robots/birds/nodes on the bounding convex hull (boundary), while ensuring that the robot maintains a safe distance from the flock, avoids flock fragmentation, and improves the cohesiveness in the bird flock. Stability and Performance of the herding strategy are studied.

Criteria for exponential stability among bird formations are determined using Graph–Theoretic approaches.

  • S. Gade, A. Paranjape, S-J. Chung. “Robotic Herding using Wavefront Algorithm: Performance and Stability”, AIAA Guidance, Navigation, and Control Conference (SciTech-16), San Diego, CA, 4-8 January, 2016. (pdf) (link)
  • S. Gade, A. Paranjape, S-J. Chung. “Herding a Flock of Birds Approaching an Airport Using an Unmanned Aerial Vehicle”, AIAA Guidance, Navigation, and Control Conference (SciTech-15), Kissimmee, FL, 5-9 January, 2015. (pdf) (link)

Collaborative Missions using Unmanned Aerial Systems

Urban search-and-rescue is considered a “multi-hazard” discipline, as it may be needed for a variety of emergencies or disasters. Use of Unmanned Aerial Systems (UAS) in urban search and rescue can provide significant help to the response teams by relaying real time information about the victim location and status, assess damage of the disaster and over and above all it can help in active planning of response.

An Ant-colony based hybrid search algorithm is presented for applications in search in an urban setting. This algorithm is scalable and robust to loss of agents. Its distributed nature makes it ideal for applications in large scale search operations. Human-in-the-Loop hierarchical control architecture is presented here for UAV flock management. On-line mission reconfiguration strategies are presented and their utility in search effectiveness and efficiency is presented.

  • S. Gade, A. Joshi. “Heterogeneous UAV Swarm System for Target Search in Adversarial Environment”, IEEE International Conference on Control, Communication and Computing (ICCC-13), Trivandrum, India, December 2013. (pdf) (link)
  • S. Gade, A. Joshi. “Human-in-Loop Hierarchical Control of Multi-UAV Systems”, International Conference on Intelligent Unmanned Systems (ICIUS-13), Jaipur, India, September 2013. (pdf)

3-P Method – A Vanishing Point Estimation Algorithm

In this work we developed a algorithm for estimation of vanishing points in structured environments. It utilizes location of 3 collinear points in image space and their distance ratio for VP estimation. We present an algebraic derivation for the proposed 3-Point (3-P) method. It provides us a non-iterative, closed-form solution. Computational time requirement for 3-P method is shown to be much less than the standard least squares based method.

  • V. Saini, S. Gade, M. Prasad, S. Chatterjee. “The 3-Point Method: A Fast, Accurate and Robust Solution to Vanishing Point Estimation”, International Conference in Central Europe on Computer Graphics, Visualization and Computer Vision (WSCG-13), Pilsen, Czech Republic, June 2013. (pdf) (link)