[Paper Review] EGO-Planner: An ESDF-free Gradient-based Local Planner for Quadrotors

Xin Zhou, Zhepei Wang+3 more|arXiv (Cornell University)|Aug 20, 2020

Robotic Path Planning Algorithms17 references9 citations

TL;DR

This paper proposes EGO-Planner, an ESDF-free gradient-based local planner for quadrotors that eliminates precomputed Euclidean Signed Distance Fields (ESDF) by directly computing obstacle gradients from trajectory collisions and a guiding path. By only extracting necessary obstacle information and using anisotropic curve fitting for refinement, it reduces computation time by over an order of magnitude while maintaining competitive trajectory smoothness, length, and energy efficiency in simulation and real-world flight tests.

ABSTRACT

Gradient-based planners are widely used for quadrotor local planning, in which a Euclidean Signed Distance Field (ESDF) is crucial for evaluating gradient magnitude and direction. Nevertheless, computing such a field has much redundancy since the trajectory optimization procedure only covers a very limited subspace of the ESDF updating range. In this paper, an ESDF-free gradient-based planning framework is proposed, which significantly reduces computation time. The main improvement is that the collision term in the penalty function is formulated by comparing the colliding trajectory with a collision-free guiding path. The resulting obstacle information will be stored only if the trajectory hits new obstacles, making the planner only extract necessary obstacle information. Then, we lengthen the time allocation if dynamical feasibility is violated. An anisotropic curve fitting algorithm is introduced to adjust higher-order derivatives of the trajectory while maintaining the original shape. Benchmark comparisons and real-world experiments verify its robustness and high-performance. The source code is released as ROS packages.

Motivation & Objective

• To address the computational bottleneck of ESDF maintenance in gradient-based quadrotor planners.
• To eliminate reliance on pre-built ESDF maps by formulating collision costs through direct trajectory-obstacle comparison.
• To develop a lightweight, robust local planner suitable for resource-constrained autonomous quadrotors.
• To improve trajectory feasibility and smoothness through anisotropic curve fitting during post-refinement.
• To validate the planner's performance in both simulated and real-world environments with dynamic and constrained flight conditions.

Proposed method

• The planner formulates collision cost by comparing the current trajectory with a collision-free guiding path, enabling direct gradient computation from obstacle intersections without ESDF.
• Gradient forces are projected onto the colliding trajectory to push it out of obstacles, with computation only triggered when new obstacles are hit.
• A post-refinement procedure reallocates time allocation when dynamical feasibility is violated, generating a new B-spline that preserves shape while improving feasibility.
• Anisotropic error penalization is applied in curve fitting, with higher penalties on radial deviations to maintain trajectory integrity and smoothness.
• The method avoids ESDF updates entirely by only extracting obstacle information when necessary, significantly reducing redundant computation.
• The framework is integrated into a fully autonomous quadrotor system and released as ROS packages for community use.

Experimental results

Research questions

• RQ1 Can gradient-based local planning for quadrotors be achieved without pre-computed ESDF maps?
• RQ2 What is the impact of eliminating ESDF computation on planning efficiency and trajectory quality?
• RQ3 How can obstacle gradient information be effectively extracted without ESDF for real-time trajectory optimization?
• RQ4 Can anisotropic curve fitting improve trajectory smoothness and feasibility while preserving shape?
• RQ5 How does the proposed method perform in real-world, dynamic, and vision-limited environments?

Key findings

• EGO-Planner reduces total planning time to 0.81ms on average, a reduction of over an order of magnitude compared to ESDF-based methods.
• The method achieves a flight time of 24.38s, trajectory length of 42.24m, and energy cost of 196.64, outperforming Fast-Planner in speed and length while being comparable in energy efficiency.
• In dense environments, EGO-Planner generates smoother and more aggressive trajectories than EWOK, which suffers from unstable convergence due to exponential terms in its objective function.
• Real-world experiments show the drone can fly at speeds above 3.56m/s in indoor cluttered environments and over 3m/s in outdoor forests, even with dynamic wind-induced obstacles.
• The planner successfully handles sudden goal changes and limited camera FOV, demonstrating robustness in real-time, vision-constrained navigation.
• The absence of ESDF computation eliminates the dominant 70% processing overhead seen in prior methods, making the planner suitable for embedded platforms.