It's possible to use ultrasonic data to generate a cost map?

Hi, i would like to know if it's possible to use ultrasonic data from the model sensor in driving scenario to create a cost map, i'm developing a simulation for autonomous parking and i need a cost map to create motion planing function, like RRT*. I've seen the SLAM algorithm in matlab, but it accept only lidar data. It's possible to make the SLAM work with ultrasonic data? Or it's better chose another approach?

Answer 1

[Sandeep]

Hi Luan,

It is great to know that you are working on developing a simulation for autonomous parking. It is possible to use ultrasonic data for creating a cost map in a driving scenario, but it may not be the most suitable approach for your specific use case. Ultrasonic sensors are commonly used for short-range obstacle detection and are not typically used for SLAM (Simultaneous Localization and Mapping) algorithms.

SLAM algorithms, such as those implemented in MATLAB, are commonly designed to work with lidar data, which provides more accurate and detailed 3D point cloud information. Lidar sensors are better suited for creating a precise representation of the environment, which is crucial for motion planning in autonomous systems.

To create a cost map, you typically need more comprehensive data about the environment, including information about obstacles' size, shape, and position. Lidar sensors are better suited for capturing this level of detail. Therefore, it may be more appropriate to consider using lidar data or a combination of lidar and ultrasonic data to generate a cost map for your motion planning function.

If lidar data is not available or not feasible, you could explore alternative approaches such as using camera-based perception algorithms or depth sensors like time-of-flight cameras. These sensors can provide more detailed information about the environment, allowing you to create a more accurate cost map for motion planning.

Hope you find it helpful.

Answer 2

[Akshai Manchana]

Hi Luan,

What Sandeep said is correct. Usually for the use cases like autonomous parking which involves accurate state estimation it is always advised to use a better sensor like LIDAR (2D-3D) or camera. But if you have created a high resolution ultrosonic system using sensor arrays or any other novel method and you would like to use the system for localization you can follow the example.

The example show how to create rays and angles from ultrasonic data and create a dummy lidar scan from ultrasonic data. Once you create a lidarScan object from ultrasonic data you can use lidarSLAM utilities for mapping and localization. The maps can be very sparse and localization may run into issues while using this sensor. You can use better sensor suites like camera+IMU shown in example for cost affective replacement. However these sensor suites may need more computational power.

Will be happy to discuss your use-case and suggest more alternatives.

Thanks & Regards

Akshai Manchana