PROBLEM
According to the National Academy of Engineering (NAE), one of the 14 challenges to enhance life on this planet is ‘How can robots assist to explore the cosmos.’ Sending a person to Mars is expected to cost more than $100 billion [1], but deploying a robot costs merely $2.9 billion [2]. Sending an individual additionally increases the hazards to one’s safety. If a robot on an exploratory trip malfunctions, the worst-case scenario is that the robot does not return. However, if a person is involved in a botched operation, the consequences are even direr [3].
Robotic systems can perform tasks that individuals are unable to perform as well as survive severe conditions such as radiation and extreme heat. Robotic systems also do not require a continual supply of water or food and can be recharged using solar panels; some can also be custom-built to perform specific jobs and outfitted with sensing technologies to gather data [4]. Individuals undertaking such jobs may incur extraordinarily high costs since specific equipment must be created to protect them.
OBJECTIVES
Designing a portable, cost-effective robot to explore hazardous environments for the safety of humans.
2Gs (Two end goals):
1. AI navigation system
- LiDAR sensor system to detect obstacles and collect data
- Collision avoidance with reinforcement learning
- Robot Operating System (ROS) for sensor control
- SLAM for data mapping and obstacle predictions
2. Optimized structure design
- Design and 3D print a feasible chassis to hold all the components
- Optimize structure to improve durability
- Assemble all hardware components and test the enclosure to fit them
Our expected outcomes consist of the following:
- Maze navigation is accomplished by a machine learning algorithm
- An omni-wheels control system
- A lightweight chassis (Optimized topology design)
- If possible, we would also like to have ROVA navigate at various elevations and steep inclines
Education Objective:
By completing the ‘BEGINNER‘ assignment, high school or college students learn 3D printing and laser cutting skills. They will be required to design, develop and 3D print a stackable storage container for hardware components such as nuts, bolts and screws before proceeding onto the robot structural system; here, they will be required to laser-cut the 2D chassis framework.
Undergraduate students must develop their fundamental programming expertise in either, python or C, while experimenting on the Nvidia Jetson Nano. Additionally, they will also study mechanical structures by creating a three-dimensional robotic systems chassis with CAD software such as AutoCAD, onshape, SOLIDWORKS to house electronics for the robot. Another activity will be to calibrate any types of sensor, which requires basic programming, and integrating it into the microprocessor. This project can be connected to an undergraduate class known as ‘INTRODUCTION TO ROBOTICS‘.
Research Objective:
This project’s area of interests is data gathering and obstacle mapping in hazardous situations. The robotic navigation system will be enhanced via machine learning; ROVA should be capable of autonomously detecting, recording, and avoiding obstacles as ROVA may have to traverse difficult terrain with varying altitudes and recognise impediments that cannot be traversed.
SYSTEM STRUCTURE
The structure system can be divided into five sub-systems: Motion, Power Supply, Sensor, Microcontroller, and Chassis. The Optimal design focus on the chassis design and this sub-system is also the foundation for all other four sub-systems. The arrows in the picture below show the correlation between the subsystems. There are also several prototyping methods used for designing the structure. The major tools are CAD and 3D printing: use CAD to design or simulate the environment and use 3D printing to produce physical components for testing or assembly.
INITIAL DESIGN IDEAS
The initial ideas comprise two parts in two pictures (Below) respectively: the left picture indicates a general structure of the robot and what potential features may be added; the right picture focuses on how to use the omni-wheels [5], particularly how to coordinate each of the four wheels to determine movements of the robot.
The blue circle in the left picture, for example, shows a design idea of having a drill in the middle bottom of the robot to collect samples or break through obstacles. The yellow circles demonstrate different types of chassis layouts [6]. The green circle in the right picture similarly shows a design of the chassis by adding a damping system to optimize stability. The orange circles illustrate different layouts of omni-wheels based on existing products.
Ideas of base structure design and movement of Omni wheels
STRUCTURE SELECTION CRITERIA
Alternatives are ranked using a point-based system. The original plan is to adopt an existing robot structure while the group focuses on developing the software. Thus, the scoring system evaluates the robot structure through aspects such as simplicity, feasibility, and material usage.
The first two evaluating criteria are worth ten points each, while the last criterion is worth five points. If a condition is met, the structural alternative receives five points. And if it is not entirely accomplished, fewer points will be given.
1. Simplicity
The structure should be simple to construct while retaining essential characteristics. To reduce time, the production process should be as efficient as possible.
2. Feasibility
The structure should be simple to construct while retaining essential characteristics. The production process should be as efficient as possible to reduce time.
3. Efficiency of Material Usage
ROVA is required to be lightweight, and its dimensions should not be too large or it cannot be fit into small spaces. If it is compact and lightweight, this will also help to conserve resources.
DESIGN ALTERNATIVES
An example of a rover inspired by NASA’s differential-bar rovers – it features six wheels and numerous joints that allow it to traverse across various terrains; it is capable of climbing over small obstacles. The link can be found HERE.
Most of the files are pre-existing, saving plenty of design time. The robot is also easy to construct and there are only few joints. The efficiency of material use, however, is not high. The rectangular enclossure, in particular, is not shaped to fit the internal electronic components. Score for this option therefore is: 6 (Simplicity)+7 (Feasibility)+3 (Material) = 16
A robotic system comprising only 19 pieces, weighing 2.8 kg and is small in size – this robotic system is extremely adaptable to different types of terrain and is waterproof. However, the CAD files must be bought ($10). The link can be found HERE.
The robot has the most durable design among the three options, but at the cost of simplicity. The robot also has a damping system and a special structure to be waterproof. The score in simplicity therefore is only 4. Score: 4+7+4 = 15
A low-cost robot with minimal functionality – to execute certain paths, the rover features a line tracking sensor. The design is extremely simple to set up and can be programmed in a variety of ways. The link can be found HERE.
The robot can be built based on existing resources from the JetBot project including CAD files and basic codes. As the robot has already been built many times, as indicated from online reviews, the structure should be reliable and practical. The robot is also designed and programmed by Nvidia partners, and this option is given 8 for the feasibility criteria. The structure already exists and does not require prototyping, thus saving prototyping materials. Score: 8+8+4=20
In conclusion, the third alternative scored the highest of 20 points and was chosen as the final design. This robot design also comes with rudimentary code to help prevent obstacles, which further helps with programming. Nevertheless, after a review of the electronic components for this project, it is clear that the design structure cannot be directly used due to the unique combination of these components. The chassis must be specially designed to fit the components such as the motor driver, power supplies, LiDAR, ultrasonic sensor, line tracking sensor, and a microcontroller.
TIMELINE
LEAN CANVAS
A lean canvas breaks down an aim into several business segments. All the segments are tailored to meet customer demands and achieve company growth. Starting from problem and solution statements, the lean canvas shows a basic financial analysis in the cost structure and revenue stream segments. The lean canvas also demonstrates the uniqueness of the product in the unfair advantage. The Customer Segments shows the targeted end-users of the product and the Channels aim to deliver products to end-users in time.
Problem Humans cannot perform certain exploratory tasks and safety risks and costs are too high for humans to work in severe conditions | Solution A wheeled, remotely controlled robot adaptable for different terrains | Unique Value Proposition ‘Mission impossible, not for robots’ | Unfair Advantage Multiple patentable structure design and sensor technology | Customer Segments •High-tech companies developing products to work in severe conditions (E.g., High pressure) •Governments that sponsor exploratory projects in unknown chartered | |
Key Metrics •Achieve break-even point within 5 years •50% profit margin •P/E ratio of 30 | Channels •TV, radio, and magazine ads •Endorsement by senior advocacy groups •Showroom and trade show | ||||
Cost Structure •Manufacturing and material ~30% •R&D ~40% •Marketing and sales ~25% | Revenue Stream •Recurring revenue in maintenance via contracts with high-tech companies •Transaction revenue from retailers & internet •Licensing revenue (Patents) |
STRUCTURE DECOMPOSITION
The pictures below demonstrate different layers of the robot. They also demonstrate the location of various components on each layer. There are in total three layers, with the top layer having an extra layer for the LiDAR. Each layer is designed based on the electronic components. The structure of this robot is special as a robot with such a combination of components is unique and existing chassis cannot be used. (Click the images below for an expanded view)
PROEJECT VISION
This project may be separated into four sections based on its intricacy. Each section is interconnected with the others, and the individual may select which section to begin with depending on their skillset. Our recommendation is that High school and College students should begin with the “BEGINNER” project; Undergraduates should begin with the “INTERMEDIATE” project; and Master’s students may begin with the “ADVANCED” project but should consider the “INTERMEDIATE” project as it covers robotic hardware. The “EXPERT” project is primarily intended for PhD students.
Task 1: Design and produce a small shelf bin to store hardware components such as screws, bolts, and other parts. The container should have the functionality to be layered. The dimensions should be approximately 160 x 80 x 36 mm and dividers for different-sized bolts should be included in the shelf bin. This tutorial will be helpful to learn 3D printing quickly. Once you are familiar with the design process, you can create containers for other electronic components and categorize them with dividers.
Task 2: 3D print two shelf bins and test whether they can be mounted. Note that the two containers may require a clearance of up to 0.5 mm in order to be stacked.
Bonus task: Create a 2D chassis for this robot. Start by downloading the 3D files of the electronic components and adjusting the dimensions in CAD software based on the sizes of the components. The goal is to provide physical support for components such as the motors and motor driver. The challenge would be to locate the holes for the bolts and screws. Once the design is complete, cut it out with a laser cutter using the correct settings. Here is a Tutorial for laser cutting.
Example CAD files are here: 1) Shelf bin 2) 2D Chassis. Notice I used Fusion 360 to design all the structures and the files are in .f3d format. Your designs do not have to be the same as mine but they should meet the aims listed above.
Task 3: Construct the Elegoo Robot Kit using the instructions; test and identify changes required in order to enhance the robot as shown here.
Task 1: Create a 3D chassis for a wheeled robot using a LiDAR sensor and Nvidia Jetson Nano microcontroller. It is worth noting that the robot link in the “BEGINNER” project uses an Arduino microcontroller rather than an Nvidia Jetson Nano. As a result, the design needs to be modified for the new microcontroller.
The CAD files for the main components are available here: LiDAR and Jetson Nvidia Nano Microcontroller. An additional power bank will be used to operate the Nvidia microprocessor solely due to the high voltage demand of the microprocessor. The original battery from the ELEGOO robot kit will be used to power the motor driver. The additional power bank has dimensions of 140 x 68 x 15 mm. You may scroll up to the STRUCTURE DECOMPOSITION section to view the arrangement of the components. However, this is only one version and you can relocate the components to improve space efficiency.
The chassis file can be found here. Designing the bottom layer chassis is the most challenging part of the prototyping. You may start with the existing chassis from the JetBot and use the dimensions of the motor enclosure, or you can design on your own. Using the existing chassis will save time from designing the motor enclosure but will also cost much time removing unnecessary features, especially because the JetBot chassis is for a three-wheeled robot but we have a four-wheeled one. Another drawback is that the JetBot chassis does not have a design timeline and is a .stl file, making it difficult to remove or modify a feature. Thus, there is a tradeoff between using existing chassis or designing your own.
To make the prototyping process simpler, I made two tutorial websites for students to learn quickly: general prototyping and prototyping for chassis layers. There are also tips in each tutorials and these are the lessons I learnt usually from failures. They aim to provide students with a better starting point in prototyping and save time for students.
In this section, students will be guided to install ROS (Robot Operating Software). Here is some information about ROS. There is also a ROS tutorial that is very helpful. The tutorial is a bit long but you can select the key lessons to learn.
Bonus task: explore Rviz and Gazebo which are integrated software within ROS; these allow us to simulate prototypes using various variables such as joint variables, etc. We will be initiating this section using a sample “Armbot” project.
Through this section, LiDAR will be explored using its default program “frame_grabber” to grab CURRENT data, as well as using LiDAR in ROS as shown in the videos below.
Video 1.1. This video shows LiDAR working alongside ROS to show a physical representation of the scanned map in real time.
Video 1.2. This video shows the raw data being scanned from the LiDAR via the terminal on the Jetson Nano.
Initially, this section will be integrating LiDAR onto ROS through Hector to give a clearer map which will then be used in a further simulation using the link provided here.
After creating a URDF file for ROVA, as well as optimizing all variables in Gazebo, we will be seeing how to integrate LiDAR within the simulation.
Post-simulation will be assessing the motion planning API, which will create a predictive planning scene allowing the operator to simulate collision avoidance. This was done by utilizing and enhancing the “Jetbot” GitHub repository provided by Nvidia.
Bonus: After optimizing ROS and Hector for an offline map, you can control ROVA real-time by connecting the Jetson nano through its IP address. This can be done by using the Gazebo model that we have created to act as a placeholder in Hector and SLAM. Whilst we move ROVA through the keyboard wirelessly, the LiDAR scans the data in real-time, slowly building a map on Hector.
TUTORIAL RESOURCE
Click the following images to view tutorials. Each tutorial is followed by a Youtube video in the end. The tutorials aim to be precise and help students to quickly develop their hand-on skills. It is strongly suggested that students practice the skill immediately after reading the tutorials to solidify their knowledge. And if students find anything confusing, they can go back and reread the parts of the tutorials for solutions. There are four tutorials: 1) Robot structure prototyping 2) Robot components 3) Laser cutting tutorial 4) 3D printing tutorial
FINAL PRESENTATION
Here is our final presentation! In this video, we introduced how we decide to design this robot, what we did, and what future directions can be. The tasks are divided into two parts for two people: Optimal Design and Smart AI System. Future works can be based on either of these two parts, or a combination of these two. The major tasks for future works include redesigning the chassis to improve space efficiency and strength, programming the robot to avoid obstacles at a slope and the robot can memorize the location of the obstacles, and integrating Arduino, LiDAR and Jetson Nano for the robot (Using two microcontrollers).
REFERENCES
[1] NAE GRAND CHALLENGES (2019). Engineer the Tools of Scientific Discovery. Grand Challenges – Engineer the Tools of Scientific Discovery. http://www.engineeringchallenges.org/8965.aspx.
[2] Swanson, S. (2019). Are astronauts worth tens of billions of dollars in extra costs to go to Mars? THE CONVERSATION. https://theconversation.com/are-astronauts-worth-tens-of-billions-of-dollars-in-extra-costs-to-go-to-mars-111348
[3] NASA Science (2021). Why Do We Send Robots To Space? NASA Science space place. https://spaceplace.nasa.gov/space-robots/en/
[4] Anderson, G. T., Tunstel, E. W., & Wilson, E. W. (2007). Robot system to search for signs of life on mars. IEEE Aerospace and Electronic Systems Magazine, 22(12), 23-30.
[5] Babjak, J., Novák, P., Kot, T., Moczulski, W., Adamczyk, M., & Panfil, W. (2016, May). Control system of a mobile robot for coal mines. In 2016 17th International Carpathian Control Conference (ICCC) (pp. 17-20). IEEE.
[6] Tzafestas, S. G. (2013). Introduction to mobile robot control. Elsevier.