Path Planning and Control for Autonomous Vehicles

Path Planning and Control for Autonomous Vehicles

By Brandon Taylor, Marketing Manager, AUTONOMOUS SOLUTIONS, Inc.

 

 

Autonomous vehicles are gaining traction across multiple industries, but the concept of self-navigation continues to raise questions. Without a human directly controlling the navigation, how does a vehicle know where to go, how to get there, or what real-time decisions need to be made along the way?

Accurate and reliable path planning is paramount to a safe and successful operation of an autonomous vehicle.

Also, the ability of the vehicle’s sensors to detect obstacles is a critical element of collision avoidance. Varying shapes, sizes, and locations of obstacles can create unique challenges for vehicle sensors. Some conditions, such as heavy snowfall, can cause a false-positive or false-negative reaction when there is no impending collision.

Precision control algorithms, developed by ASI, use artificial intelligence (AI) and machine learning to enable autonomous vehicles to accomplish their tasks safely and efficiently.

There is a lot of buzz about, and varying definitions for, both artificial intelligence and machine learning. Put, artificial intelligence is the capability of a machine, or in this case, a vehicle, to imitate intelligent human behavior. Machine learning is an application of artificial intelligence that allows a system to learn and improve from experience without being explicitly programmed automatically. Essentially, it enables a computer to determine for itself, using the data it gathers.

Some examples of machine learning and artificial intelligence:

Auto Tuning – By monitoring and comparing actual and desired behaviors, parameters are automatically adjusted for continuous improvement.

Obstacle Identification and Classification – Enhancements in world modeling and object learning are shared between vehicles for collaborative and improved behaviors.

Convoy Operations – Convoy solutions increase the safety and efficiency of vehicle groups to travel through unknown and harsh terrain.

Path Planning and Control

The basic framework of path planning and control starts with programming an objective for the autonomous vehicle to achieve. To accomplish this task, the machine must choose a path and adjust to obstacles, terrain, and changing conditions to reach its destination safely.

The vehicle can achieve the desired path by continually monitoring the current vehicle state (position and heading) and comparing it to the desired vehicle state. A LiDAR-based perception system is used to ensure the vehicle does not collide with anything along the path. This is accomplished by creating a cost surface that penalizes getting close to objects and penalizes being off the desired track. The future trajectory of the vehicle is calculated by finding the path of the vehicle that minimizes the cost surface through the environment. A nonlinear optimization algorithm uses the vehicle kinematics to find the minimum-cost, possible way.

Figure 1 shows an example of how an operator might build a path, and plan actions for the vehicle to execute, using ASI’s command and control software, called Mobius.

Step 1 shows the Path Builder is used to edit the curves of the desired path. The user can place the way anywhere in the map. This path could also be created by recording the vehicle while it is driven in the desired location.

Step 2 shows the event planning. This allows the user to specify specific actions that the vehicle should perform along the path. These actions can be customized by the user to do things such as stop and wait, accelerate, save a photo of the surrounding area, or perform other actions of which the vehicle is capable.

Step 3 shows the assignment of the vehicle to the planned path. This is as simple as selecting the vehicle and then picking away for that vehicle.

Autonomous Navigation

ASI’s AI algorithms are used to facilitate safe and reliable navigation of unknown or dangerous terrain to arrive at the desired location. A terrain model can be computed onboard the vehicle by fusing sensor data from LiDAR, camera, and RADAR. With this terrain model, the vehicle can predict future behaviors for hazard avoidance and optimal trajectory selection.

World Modeling

External sensors on the vehicle collect real-time data from its surroundings, which is then processed by a sophisticated algorithm to build an accurate model of the environment. This model can be accessed and updated by any autonomous vehicle entering the area.

Leader-Follower Operation

Another autonomous vehicle feature is the ability to follow the leader. Whether the lead vehicle is manned, unmanned, or even a person on foot, the system enables autonomous vehicles to support the defined leader while still maintaining their collision avoidance and terrain navigation features.

GPS-Denied Navigation for Manufacturing Facilities

In addition to outdoor ground vehicles, efficient path planning and control also applies to autonomous indoor solutions, such as ASI’s proprietary GPS-denied navigation solution for indoor vehicle positioning. This solution is ideal where satellite-based navigation is not an option, such as inside manufacturing facilities and shipping warehouses.

Historically, indoor positioning has relied on infrastructure installations such as magnets or wires in the floor, or laser reflectors around a perimeter. These methods only allow travel on installed routes or work only in conditions where nothing is blocking a vehicle’s line of sight to the reflectors.

This GPS-denied navigation technology can be applied in multiple real-world operations, including the following examples:

Autonomous floor-cleaning robots operating in a retail shopping environment where shelving remains the same, but smaller displays may be moved around. AGVs are transporting inventory through a stock room or warehouse areas.

Security robots following routes that involve movement in and out of buildings. Material handling is a facility with frequent, unexpected change, providing valuable system uptime, flexibility, and ease of deployment.

As autonomous vehicle technology improves, and the demand for these solutions continues to grow, advanced path planning and control will become even more critical to the success of manufacturers. Autonomous Solutions, Inc. remains an industry leader as it pushes advancements in this area. Global companies have used ASI’s solutions across industries in material handling, agriculture, automotive, mining, military, and security. ASI customers have experienced enhanced safety, increased efficiency, and improved accuracy, as the company furthers its mission to help organizations reach their potential through innovative robotic solutions.

Check out: Top Drone Technology Companies

Weekly Brief

Top 10 Industrial IoT Solution Companies - 2019

Read Also

"A Strategy to Differentiate with Additive Manufacturing"

Bill Massaro, Director of Advanced Manufacturing, Moog, Space and Defense Group [NYSE:MOG.A]
Data Data Everywhere, but Not a Bite to Eat

Data Data Everywhere, but Not a Bite to Eat

Joe Cichon, Vice President Manufacturing Technology at INX International Ink Co.
New Tools Enable AM to Create Differentiating Aerospace Products that Make the World Better

New Tools Enable AM to Create Differentiating Aerospace Products that Make the World Better

Melissa Orme, Vice President, Boeing Additive Manufacturing, The Boeing Company [NYSE:BA]
Leveraging the Human Connection for Real IOT Conversation

Leveraging the Human Connection for Real IOT Conversation

Paul McKeithan, Head of Digital Services at Bühler Group
Evolving Past Robotic Ambition to Adoption

Evolving Past Robotic Ambition to Adoption

Kevin Paramore, Emerging Technology Commercialization Manager, Yale Materials Handling Corporation
How the DFM Improve competitiveness

How the DFM Improve competitiveness

Alberto Flores Jiménez, Global Head and Director of Multidoor Architecture and Ice and water Systems, Electrolux [STO:ELUX-B]