Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

E-MPC: Edge-assisted Model Predictive Control

Yuan-Yao Lou, Jonathan Spencer, Kwang Taik Kim, Mung Chiang

TL;DR

This work proposes a novel framework for edge-assisted MPC (E-MPC) for path planning that exploits the heterogeneity of edge networks in three important ways: varying computational capacity, localized sensor information, and 3) localized observation histories.

Abstract

Model predictive control (MPC) has become the de facto standard action space for local planning and learning-based control in many continuous robotic control tasks, including autonomous driving. MPC solves a long-horizon cost optimization as a series of short-horizon optimizations based on a global planner-supplied reference path. The primary challenge in MPC, however, is that the computational budget for re-planning has a hard limit, which frequently inhibits exact optimization. Modern edge networks provide low-latency communication and heterogeneous properties that can be especially beneficial in this situation. We propose a novel framework for edge-assisted MPC (E-MPC) for path planning that exploits the heterogeneity of edge networks in three important ways: 1) varying computational capacity, 2) localized sensor information, and 3) localized observation histories. Theoretical analysis and extensive simulations are undertaken to demonstrate quantitatively the benefits of E-MPC in various scenarios, including maps, channel dynamics, and availability and density of edge nodes. The results confirm that E-MPC has the potential to reduce costs by a greater percentage than standard MPC does.

E-MPC: Edge-assisted Model Predictive Control

TL;DR

This work proposes a novel framework for edge-assisted MPC (E-MPC) for path planning that exploits the heterogeneity of edge networks in three important ways: varying computational capacity, localized sensor information, and 3) localized observation histories.

Abstract

Model predictive control (MPC) has become the de facto standard action space for local planning and learning-based control in many continuous robotic control tasks, including autonomous driving. MPC solves a long-horizon cost optimization as a series of short-horizon optimizations based on a global planner-supplied reference path. The primary challenge in MPC, however, is that the computational budget for re-planning has a hard limit, which frequently inhibits exact optimization. Modern edge networks provide low-latency communication and heterogeneous properties that can be especially beneficial in this situation. We propose a novel framework for edge-assisted MPC (E-MPC) for path planning that exploits the heterogeneity of edge networks in three important ways: 1) varying computational capacity, 2) localized sensor information, and 3) localized observation histories. Theoretical analysis and extensive simulations are undertaken to demonstrate quantitatively the benefits of E-MPC in various scenarios, including maps, channel dynamics, and availability and density of edge nodes. The results confirm that E-MPC has the potential to reduce costs by a greater percentage than standard MPC does.
Paper Structure (39 sections, 20 equations, 7 figures, 1 table)

This paper contains 39 sections, 20 equations, 7 figures, 1 table.

Table of Contents

  1. Introduction
  2. Motivations and Challenges
  3. Related Work
  4. Outline and Summary of Contributions
  5. Problem Statement
  6. MPC Local Planning Cost-Minimization Problem
  7. Edge-Assisted Model Predictive Control (E-MPC)
  8. Increased Computational Capacity
  9. Localized Sensor Data
  10. Localized Observation Histories
  11. System Framework
  12. Experimental Simulation
  13. Motion-primitive library
  14. Global reference path by RRT* algorithm
  15. Trajectory sampling and cost evaluation
  16. ...and 24 more sections

Figures (7)

  • Figure 1: The examples of how edge nodes can assist MPC. Under assistance from one or multiple edge servers, the agent precisely tracks the dotted reference path, avoids unknown environment changes, and deviates from the reference path for a shortcut based on historical data.
  • Figure 2: The high-level overview of E-MPC system framework: the black boxes are the obstacles in the surrounding areas, while the costly areas represent the agent's uncharted environment.
  • Figure 3: Motion primitive dictionary.
  • Figure 5: Evaluation bar charts for the experiments: (a) Increased computational capacity (b) E2E latency impact (c) Density of edge servers (d) Availability of edge servers (e) Localized sensing information: Icy road (f) Local sensing information: Mud area.
  • Figure 6: Heat maps for the experiment localized observation histories evaluated using nine weight parameter $\beta$ and nine cost-optimal prior libraries.
  • ...and 2 more figures