BEGIN:VCALENDAR VERSION:2.0 PRODID:-//wp-events-plugin.com//7.2.3.1//EN TZID:Asia/Kolkata X-WR-TIMEZONE:Asia/Kolkata BEGIN:VEVENT UID:123@cds.iisc.ac.in DTSTART;TZID=Asia/Kolkata:20250428T160000 DTEND;TZID=Asia/Kolkata:20250428T170000 DTSTAMP:20250421T124312Z URL:https://cds.iisc.ac.in/events/ph-d-thesis-colloquium-102-cds-28-april- 2025-scalable-platform-for-intelligent-orchestration-of-autonomous-systems -across-edge-cloud-continuum/ SUMMARY:Ph.D: Thesis Colloquium: 102 : CDS: 28\, April 2025 "Scalable Platf orm for Intelligent Orchestration of Autonomous Systems Across Edge-Cloud Continuum" DESCRIPTION:DEPARTMENT OF COMPUTATIONAL AND DATA SCIENCES\nPh.D. Thesis Col loquium\n\n\n\nSpeaker : Ms. Suman Raj\nS.R. Number : 06-18-01-10-12-20-1- 18450\nTitle :"Scalable Platform for Intelligent Orchestration of Autonomo us Systems Across Edge-Cloud Continuum"\nResearch Supervisor: Prof. Yogesh Simmhan\nDate &\; Time : April 28\, 2025 (Monday)\, 04:00 PM\nVenue : CDS # 102\n\n\n\nABSTRACT\nThe benefits of autonomous mobile platforms\, s uch as unmanned aerial vehicles (UAVs) equipped with onboard cameras\, are enhanced by compact edge accelerators that are co-located\, such as the N VIDIA Jetson with 100s of CUDA cores. They enable rapid inference of Deep Neural Network (DNN) models and computer vision algorithms to support real -time analytics workflows for diverse domains\, ranging from smart crop mo nitoring to assisting Visually Impaired People (VIPs)\, either individuall y or as part of a fleet.\n\nHowever\, programming such drones and edge dev ices for efficient\, resilient and responsive operations poses challenges. The limited compute capacity of edge devices needs to be intelligently co mplemented by cloud computing to offload compute-intensive analytics in a timely and cost-effective manner. Routing fleets of drones to accomplish c omplex tasks requires us to optimize for and adapt to network variability\ , edge failures\, latency and energy constraints\, and monetary costs. Fur ther\, these need to be intuitively programmable to design practical appli cations across distributed autonomous platforms and edge resources. We add ress these challenges in this dissertation.\n\nFirst\, we design a task sc heduling strategy\, GEMS\, that performs real-time decisions for DNN infer ence tasks generated by drones\, to execute them either on local edge acce lerators or remote cloud resources. The goals are to maximize the Quality of Service (QoS) and Quality of Experience (QoE) for a VIP assistive appli cation within the deadline constraints of each task. GEMS accounts for the task deadline\, cloud and edge pricing\, and dynamic network variability. Our realistic experiments using up to 84 emulated drones show up to 2.7x higher QoS utility\, up to 75% higher QoE utility\, within ±1m/s3 Jerk\, up to 42% lower yaw error\, and a task completion rate of up to 88% compar ed to state-of-the-art baselines for diverse computer vision workloads.\n\ nNext\, we study co-scheduling of DNN inferencing and routes for a fleet o f drones used in collaborative applications such as smart agriculture. The drones need to visit a set of waypoints to collect data and perform analy tics\, have access to onboard edge compute\, stationary fogs at cell tower s\, and mobile fogs on public buses. We define this as a Mission Schedulin g Problem\, which is NP-complete\, and design MARC as a divide and assign heuristic to solve this optimization problem. Our simulation-based evaluat ion of MARC with fleets of up to 50 drones\nachieves a 100% task completio n rate and up to 31% higher average utility than contemporary baselines\, and is within 75% of the optimal solution solved using MILP\, which is tra ctable only for small inputs.\n\nFurther\, we explore resilient scheduling of autonomous systems to ensure continuity of service despite drone and e dge failures in the context of wildfire response\, where a heterogeneous U AV fleet helps detect stranded individuals and generate evacuation routes to safety. We develop the AeroResQ platform with algorithms that dynamical ly adapt to failures by using heartbeats across drones and an onboard dist ributed datastore. Strategies like a load balancing algorithm to address t he active requests being processed by failed drones and re-assignment of s patial regions across the available drones ensure uninterrupted detection\ , route planning\, and monitoring. Our evaluations\, conducted using an em ulated environment based on recent Southern California wildfire data\, dem onstrate the robustness of our platform under failure scenarios and fleet configurations. The system achieves real-time performance with ≤1s end-t o-end latency per evacuation request\, much below the 2s request interval\ , while maintaining over 98% successful task reassignment and completion.\ n\nFinally\, we design the Ocularone platform as an integrated Drones-as-a -Service (DaaS) programming framework that enables rapid development of an alytics-driven UAV applications across the edge-cloud continuum. It abstra cts drone navigation\, control\, and sensing into intuitive\, composable P ython-based interfaces\, and can embed the scheduling strategies that we h ave developed\, such as GEMS. This is validated for an assistive applicati on to help navigate visually impaired individuals using buddy drones. This is further enabled by accurate DNN-based distance estimation algorithms w e have developed to assess nearby obstacles\, which are fine-tuned with cu rated datasets to improve real-world accuracy. The application can be impl emented in under 40 lines of code using Ocularone.\n\n\n\nALL ARE WELCOME CATEGORIES:Events,Ph.D. Thesis Colloquium END:VEVENT BEGIN:VTIMEZONE TZID:Asia/Kolkata X-LIC-LOCATION:Asia/Kolkata BEGIN:STANDARD DTSTART:20240428T160000 TZOFFSETFROM:+0530 TZOFFSETTO:+0530 TZNAME:IST END:STANDARD END:VTIMEZONE END:VCALENDAR