Download presentation
Presentation is loading. Please wait.
1
Wireless Autonomous Trolley (WAT)
Critical Design Review
2
Randy Baron Responsible for: Image Processing Hardware Interfacing
Camera Implementation Motor Implementation Infrared Proximity Sensor Implementation
3
Phillip Hoang Responsible for: Hardware Design
Internet Communication Implementation
4
Functionality and Motivation
Go to locations specified by a user wirelessly via a web page Would synergize with a wireless vending machine; you would be able to get food from another location without standing up Possible uses Transporting food/goods at home, or in a hospital or restaurant
5
List of Key Parts and Components
Altera DE0 Nano Chassis with 4 motors ROB Motor Driver → Motor Driver Controller uCAM-II UART camera → Camera UART Module Infrared Proximity Sensor → Infrared Controller ESP8266 Wifi module → Wifi UART Module Buzzer Batteries
6
Hardware Design
7
Software Design
8
The algorithm //Inside the for loop going over the row, looks for red regions if ((r > REDREG_RED) && (b < REDREG_BLU) && (g < REDREG_GRN)){ temp_row_count++; temp_x = x; region_found = TRUE; regions.pixels_detected++; } ………………… //Outside the for loop going over the row if (region_found){ regions.cols_traversed++; regions.y = y; if (temp_row_count > regions.rows_traversed){ regions.rows_traversed = temp_row_count; regions.x = temp_x; //Go onto the next row Most image processing algorithms we looked at gave qualitative results Problem is, we needed to know pixel locations for the WAT so we could pass that info on to the motors The best solution is therefore also the simplest one; iterate over each pixel and get the color delta between one pixel and the next The snippet to the right is a sample of the process, where with a bit of math, the approximate center of a region of color (in this case red) can be found. This information will be used to decide what direction the WAT will turn
9
Our major bottleneck for the design is expected to be the UART serial connection that the camera uses. Supporting up ~115kbps we’re expecting transmission times of several seconds once we start working with real images and that’s ignoring any additional time spent properly placing the image into the SDRAM The next few slides will show some sample images thrown together in GIMP and compressed to various values alongside some benchmarks. Obviously these will not mimic real world cases but this was done to ensure we are not recompressing an already compressed image. NOTE: The processing benchmarks were collected on an Intel Pentium N3710, a 1.1GHz quad core CPU. We are in the process of porting the needed libraries to the DE0 but for now, adjusting for the performance of the Nios II used in the DE0 (~100MHz) gives a single core performance ratio of ~10:1
10
Quality of 100, worst case scenario for performance
Resolution: 640 x 480 File size: 109kB Expected transfer time: 109 / 115 = ~1 second Processing time: ~400 microseconds
11
Quality of 50, most likely case for usage
Resolution: 640 x 480 File size: 6kB Expected transfer time: 6 / 115 = ~50 milliseconds Processing time: ~200 microseconds Note how minimal the quality loss is compared to the very large performance gains
12
Quality of 10, worst case scenario for quality
Resolution: 640 x 480 File size: 5kB Expected transfer time: 5 / 115 = ~50 millisecond Processing time: 200 microseconds Performance gains are extremely minimal compared to the quality loss as compared to the 50 quality
13
Interfacing Test Plan Attach components to DE0 Nano before assembling robot Test every component individually for proper operation with hard coded test instructions Fill in test parameters (eg. Pulse duration required to move x cm) Assemble robot Run robot with test parameters Adjust test parameters until margin of error is acceptable
14
Software test plan Write code to run as a local application
Use the terminal as the primary means of comparing expected and calculated data Port design to the Nios II processor Use the LEDs and buzzer as means of comparing expected and calculated data
15
Biggest Challenges Manpower Deadlines
Half a team Deadlines Specification Report due 3 days after lab 2 Differences between Unix/Windows and the FPGA JPEG library used for test cases will likely need to be manually ported Not having hardware Tough to make a robot when you have no parts available
16
Features That Can Be Removed
Wifi Communication Destination loading would instead be done via USB or physical buttons/switches or if even those fail, entirely pre programmed Image Processing Robot would instead use dead reckoning to navigate a pre programmed course
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.