Development of a Ground Robot with a Simultaneous ...

1 Development of a Ground Robot with a Simultaneous localization and Mapping (SLAM) Capability Nikki Lopez, ASU, Mechanical Engineering Advisor: Dr. Rodriguez, ASU, Professor of Electrical Engineering, Goal: The proposed (second) FURI will build on the work of the first Spring 2018 FURI in order to develop a research-grade Ground Robot with a Simultaneous localization and mapping (SLAM) capability. Such a vehicle can be used for a multitude of applications: search, rescue, reconnaissance, surveillance, distributed sensing and communication, mapping, area/infrastructure/agriculture inspection, law enforcement, first responder assistance, etc.

2 Objective: With the above goal, the main objective of the proposed FURI 2 research is to develop and demonstrate a research-grade vehicle that will serve as a model for the Development of future additional vehicles. To do this, we intend to fully leverage our FURI 1 work. Work Accomplished During FURI 1 (Spring 2018): During FURI 1, I was able to significantly deepen my understanding of modern sensing, computing, and the use of model-based algorithms. A differential-drive (Turtlebot) Robot was developed based on the prior work of Dr.

3 Rodriguez students [6]-[10], [11]-[12]. The vehicle contained a styrene-based laser cut chassis, an Arduino microcontroller for inner-loop (cruise/position/directional) control, a Raspberry Pi for outer-loop ( simple image processing) control, an inertial measurement unit (IMU) with accelerometers and gyroscopes, an HC-SR04 ultrasonic sensor to detect obstacles and obtain range information, a Raspberry Pi 5MP (30 Hz) camera, ultrasonic sensors for range information, a 2D LiDAR (10 Hz, light detection and ranging)

4 For ranging and mapping, a spread spectrum wireless communications system, 1 Arduino motor shield, 2 motors with built-in high resolution encoders, a Lithium Polymer 12V battery. Taken collectively, the above components permitted us to begin our SLAM journey. CONTROL: Vehicle modeling and control was based on [6]-[13]. Proportional-plus-integral-plus-derivati ve (PID) controllers were implemented within the Arduino for inner-loop (speed/direction/position) control. This utilized the encoders and IMU.

5 A PID controller was also implemented within the Raspberry Pi for outer-loop image-processing based control ( following a path, obstacle avoidance) [9]. MAPPING: The onboard LiDAR was used for mapping an unknown environment (in principle: with an a priori specified accuracy in minimum time). By obtaining a map of an area, we could return to the area and navigate within the area. As the map was generated, the data gathered (a subarea map) was used to navigate within the subarea. The 2D LiDAR to construct 2D maps; near Ground -level area maps.

6 SLAM: The algorithms within [1]-[5] were used in order to simultaneously determine where the Robot was and to construct an environmental map ( map of a room). As the LiDAR data was gathered, a state estimator was used to clean it up; to better estimate where the vehicle was and to find the distance to environmental features. Two classes of estimators were examined for this research include: (1) Kalman filter [1], [3] and (2) Rao-Blackwellised particle filter [4]. Work to be Accomplished During FURI 2 (Fall 2018): Because of the limitations of the Raspberry Pi and the camera, we examined an Jetson TX2 Nvidia board with 256 GPU cores and an RGB camera.

7 The Nvidia board permitted very high-speed high-resolution image processing (near 30 frames per second). The RGB camera permitted much better depth and color information. During FURI 2, these will be added to the Robot developed during FURI 1. These will permit us to pursue much more powerful machine learning (neural network) assisted algorithms for mapping [ZZ], object detection-and-recognition [ZZ], path planning [ZZ], mapping and SLAM [ZZ], [ZZ]. The limitations of each algorithm will be examined. Critical questions to be addressed include: (1) How can we map an environment to an a prior accuracy in minimum time?

8 (2) What are the limitations the LiDAR, RGB camera and Nvidia board? (3) When is filtering really needed? (4) When do the critical algorithms to be developed ( mapping, planning, SLAM) work and when do they break down? (5) When are more complex control algorithms needed? Final Demo and Future Work: The final demonstration will showcase a prototype Ground Robot which will demonstrate mapping, planning, obstacle detection and avoidance, and SLAM - all implemented within the Robot Operating System (ROS). All results will be documented in a final comprehensive report and on the final poster.

9 The work will be submitted for publication within the proceedings of the American Control Conference (ACC), the Frontiers in Education (FIE) and ASEE Conferences. Final Documentation: All project results including answers the aforementioned critical questions will be documented within a final comprehensive report and on the required final poster. The work will also be submitted for publication within the proceedings of the American Control Conference (ACC), the Frontiers in Education (FIE) and ASEE Conferences.

10 Career Relevance: My technical passions lie entirely within the intelligent autonomous vehicle arena. For this reason, I have selected the proposed project. This project will provide an excellent step toward my next goal to complete an MS here at ASU in the area of intelligent autonomous systems under the supervision of Dr. Rodriguez. The proposed research will help me mover toward my goal of earning a PhD. FURI Advisor: Dr. Rodriguez has worked in the area of Flexible Autonomous Machines operating in an uncertain Environment (FAME) for over 30 years.