reference
Focused Trajectory Planning for Autonomous On-Road Driving
And Baidu apollo EM planning similar
autonomous vehicles on the road motion planning is often a challenging problem. Past efforts have made a separate city and highway environmental solutions. We determined that the previously Solution key advantages / disadvantages and to propose a novel two-step motion planning system
city and highway driving within a single framework. Reference trajectory planning (I) the use of intensive grid sampling and optimization techniques to generate easy to adjust and consider the human reference trajectory, obstacles and advanced instruction road geometry. Sampling the reference trajectory through the focusing, tracking trajectory planning (II) generate, evaluate, and track selection parameter further satisfies the dynamic constraints of execution. The described method retains much of the performance benefits of a detailed time and space planners to reduce the calculation
related research
path generation scheme is the basis for motion planning. Arc [1], Bezier curves [2], B-spline [3] and the five curves splines [4] has been proposed as a path of the original type. However, [1] [2] of discontinuity of curvature of disadvantages [3] [4] in the absence of visual reality makes them parameterization
of passenger planning less attractive. [5] and [6] proposed a real-time path planning algorithm
using high-order polynomial equations smooth lane change. They found a second order closed path continuity solutions, but it is difficult to generate multiple paths to avoid obstacles in urban traffic.
[7] proposes to use a polynomial of curvature of the spiral, which has an intuitive and advantages of parametric computational efficiency. The method further adapt Highway Planning [8].
We propose a two-step planning framework. Reference Trajectory Planner (I) with multiple optimization technology to generate human nonparametric reference trajectory, taking the road geometry, obstacles and barriers to advanced instruction. Tracking the trajectory planner (II) to focus on space-time sampling and evaluation.
By using the trajectory parameters, it ensures the continuity of analysis. Select and use an optimal trajectory execution. And [13], we think in our program to a higher level of reasoning module (behavior module) is an effective system. Two-step plan should produce the track reflecting the advanced instruction. Figure 2 shows the motion planning system we propose.
Reference Trajectory planning
reference trajectory generation non-parametric optimization plan for the road geometry, obstructions and higher-level motion commands trajectory. The application of a variety of optimization techniques. The absolute lack of track quality standards make it difficult to achieve optimal design standard. However, user preferences, gives a good subjective criteria. Therefore, it is important to develop an optimized way is to generate human-like tracks, and allows for easy adjustment to reflect personal preferences.
1. road edge detection and path generation sprinkle
Dynamic Programming
2. Non-parametric path optimization
3. Non-parametric speed rules
4. Optimal trajectory generation
Trace trace generation
details of trajectory generation parameters, focusing spatiotemporal sampling and evaluation.
1. Parametric trajectory generation
2. Parameter velocity generated
3. discrete track and evaluate
Path / speed generation scheme described above, the reference trajectory for focus spatiotemporal sampling. There are three factors that determine the sampling process: foresight and vision, and speed the end of the path endpoint.
Assessment
first check sampling trajectories explicitly check the static obstacles and ensure safety. And then evaluated, the measurement of both spatial and temporal proximity of the reference candidate trajectory. The lowest cost is selected trajectory: wherein Cspatial is cumulative offset lateral distance respect reference path, and Ctemporal is cumulative with respect to time shift the reference trajectory. The weight wspatial and wtemporal can adjust the relative importance given to space and time of intimacy.
