EcoCar3-ADAS. Project Plan. Summary. Why is This Project Important?

Transcription

1 EcoCar3-ADAS Project Plan Summary Scott Smith This project is the Advanced Driver Assistance System (ADAS) of the Senior Design for the EcoCar3. This will be an embedded system for the EcoCar3 that will have features such as lane tracking and object detection and recognition. The team consists of three seniors and one junior and is being overseen by Dr. Sudeep Pasricha, and Vipin Kumar, a graduate student. The objective of this project is to have a robust application that can identify and track the current lane as well as identify objects and hazards so they can be avoided or at least be made known to the driver. This is the first year ADAS has been in progress so there was no previous ground work. Brendan Isbell Alex Banner Jordan Tunnel Why is This Project Important? Assisting the driver is important because it will provide a safety net for inattentive and careless drivers. Not only will it give feedback via output on the screen, but it will lead to automated mechanical operations. For example, if a driver is fast approaching a potential collision and not braking, ADAS can apply the brakes automatically to avoid or at least lessen the impact.

2 2 Problem Statement Over a million people are in fatal car accidents annually. Anything to mitigate this problem is beneficial and of upmost importance. That is why we are creating and designing a tool to aid the driver on the chaotic roads. Notifying the driver of potential hazards could reduce the number of annual fatalities and make the roads safer for everyone, including pedestrians. Objectives The EcoCar3 team has assigned us tasks to complete by ~April, These objectives are the initial steps to building the ADAS tool. The objectives are as follows: 1. Sign Identification Source code Video Written submission 2. Sign Distance Estimation Source code Video Written submission 3. S32V Demonstration Source code Read CAN signals Video (live camera) Display to live monitor and tape Written submission 4. Vision Algorithm Integration Video Implement previous fall semester algorithms over Written submission 5. FreeScale Workshop Scott will be attending this workshop in Austin, TX The workshop starts January 28, 2016 The objectives cannot be described in detail because they are confidential and a part of the EcoCar3 competition.

3 3 Hardware Emulation EcoCAR3 Winter Workshop EcoCAR3 Deliverables Freescale and Odroid Implementation Camera and Testing Prepare Spring Presentation December January February March April May Not Complete Not Complete EcoCAR3 Complete Table 1: Projected timeline for spring semester 2016 Final Design Information Initially, the application was to run on a Freescale board using a camera as input. The code was to be developed with C++ and computer vision libraries from OpenCV on a Linux operating system. The board has since been delayed and our project has shifted gears slightly. The project will still use the OpenCV libraries, but the code will now be developed using Java on Android Studio. The application will run on an Android device, specifically a tablet that has already been purchased. We are using OpenCV and Matlab, which is the most current version, to provide libraries for image processing and computer vision. This module provides everything we ll need including Hough, Canny, and Threshold among other libraries for lane tracking and ORB and several others for object detection. Since, the application will be running on a mobile device, the best code to use is Java due to portability between devices. Eventually, this will be integrated into the EcoCar3. The board and camera will be an imbedded system for the EcoCar3 and will provide feedback, either visual or verbal, and even apply mechanical features such as braking. We are unsure if we will get to this stage. While we will continuously work to achieve this, it may be difficult to achieve the first year in the works. Not only will we have to have a working application to integrate into the vehicle system, the EcoCar3 project will also have to be far enough along to be ready for it. If either side is not ready for implementation, then it may be difficult to move forward. The EcoCar3 is in its second year and still has a lot of hands-on work to be done.

4 4 Project Status Table Process Currently Implemented Not implemented Feature Extraction Threshold, Gabor texture filter, Sobel, Intensity filter, Canny edge detector Color Segmentation, Filters for other technologies such as LIDAR Lane tracking Post processing: Hough filter Least squares, RANSAC or variation Straight line fitting Post processing: Ransac, kmeans, least squares Circle fitting, Clothoid curve Curved line fitting Lane Tracking and Prediction Kalman filter Rao-blackwell, Extended Kalman Object detection and Tracking Object Detection Haar cascades (vehicles) SIFT, SURF, ORB Object Tracking S32V Initial Testing Camera Algorithms to S32V Board Integration Particle filter such as those mention under Lane Tracking and Prediction Operating System, Specialized Kernel, modules Integration patterns with S32V Feature Extraction, Lane Tracking, and Object Detection All of the filters and image processing methods have been implemented on the C++ program. Not as many have been implemented on the Java version, but may not be needed since the processing power needs to be light on a mobile device. Design Constraints 1. Fast Detection of Objects. Faster recognition provides greater response time 2. Accurate and Visible Lane Detection 3. Color Detection Stoplights, stop signs, etc. 4. Visual and Audible Feedback 5. Delayed Freescale Board Limits time for implementation 6. Android Device Processing Power Mobile devices have limited processing power 7. Embedded System ADAS is an embedded system for EcoCar3 8. Linux OS

5 5 9. Usability Ease of use is necessary to maintain interest 10. Lighting Conditions Daytime vs Night 11. Weather Conditions 12. Budget Budget Our budget consists of the $200 per senior on our team plus what Dr. Pashricha provides for us. We were told that budget is not a problem for us. He has already purchased the tablet for us and will purchase the board, camera and any kind of supplemental material/hardware/software needed to complete the project. Risk Analysis 1. False readings Both identifying false objects and not identifying real objects can be detrimental to the performance and compromise any kind of demand for it Plan to Mitigate: o Write robust, high-functioning code that is also tested thoroughly to ensure all bugs and errors are detected and then corrected 2. Malfunctioning Hardware The board is a new design from Freescale and has not been sold commercially yet Plan to Mitigate: o The workshop in Massachusetts will provide information and resources to understand and use the board o Thorough testing will improve performance 3. Missing Deadlines Since this is part of the EcoCar3 project, deadlines are hard and strict. Missed deadlines could delay other EcoCar3 teams Plan to Mitigate: o Clearly layout project o Divide tasks among team members Assign tasks according to team members strengths if possible o Communicate issues Multiple eyes can increase productivity

6 6 Conclusion Laying a good foundation for programming is very important since this project will be continued by other seniors for the next few years. The programming must be efficient, robust and accurate. Otherwise, implementing it into a car will be unnecessary work and could possibly turn into a dangerous distraction. Identifying lanes on the road and potential hazards is a very useful tool if it is done right. Our goal is to make a product that is both energy efficient and a beneficial tool to ensure safer driving.