The goal of this project is to build a beach cleaning robot, capable of autonomous movement, obstacle avoidance, and trash disposal. The motivation behind this project is purely environmental. We see Honu as a cost effective means of cleaning the beach that moves away from traditional methods ranging from labor-intensive manual cleanups to expensive, large, and clunky shore-sweepers that have to be driven by specialist operators.
Together, we made up a group of four for our Columbia University senior design team. The members on this team were myself, Jamie Hall, Bradley Beeksma, and Daniel Gonzalez.
The robot is programmed to clean the beaches of trash by combing through the trash and allocating it in the receptacle inside. It uses geofencing to choose the area to roam, and goes along a set path to clean the beach. Honu also has built in obstacle avoidance by using IR sensors, micro switches, and a humidity sensor. With this hardware and the usage of dijkstra's algorithm, the robot avoids any obstacles that may impede its path and makes routes around said obstacle.
Honu consists mainly of marine plywood and an aluminum chassis. In conjunction with the sensors listed below, Honu also uses a Raspberry Pi and an Arduino to control all of its electronic operations.
I was in charge of the creation of the electronics system as well as its assembly and implementation. All of the electronics information was obtained as the project progressed more and more with little background knowledge prior.
P: Signal from the sensors could not be read by the Arduino or Raspberry Pi.
S: It was discovered by using a multimeter that there was bad continuity between the sensor and the Arduino/Raspberry Pi. This was solved by running new wires, with very few soldering points in between, from the sensor to the Arduino/Raspberry Pi.
P: Wheels would not work well in sand due to traction, and treads were too expensive to buy
S: Our team found a solution and some blueprint for wooden treads to make ourselves online. We modified this to suit our needs and 3D printed batches of treads that were connected with binding posts.
P: Creating the shell for Honu
S: Making the shell out of 3D printed plastic would be too expensive and cumbersome to put together. We also considered other options such as using carbon fiber. We also had trouble finding material for making a shell. MDF and scrap wood was too crumbly and unusable once put through a laser cutter, and hardwood was too expensive to purchase. Instead, we got plywood from the architecture department for free. Because it was too thick though, we laser cut an initial blueprint on the top of the sheet and then ran it through the bandsaw.
P: Wheel Alignment Issues. The parts holding the treads together would not align properly.
S: The sprockets for the tank treads were misaligned so the treads would roll off quickly. Mounting brackets were created in order to secure the sprocket at two points on Honu. Doing this allowed the sprockets to be aligned as well as make Honu run straighter than before.
P: Honu isn't driving straight.
S: One side of the wheels was tighter than the other due to excess JB Weld. Unfortunately, there was not a mechanical fix to this. Instead, this was fixed by adding a bit more power to one side for a short period in order to make Honu drive straight.
P: The ocean tide might sweep Honu away
S: Pull up tide prediction from weather reports on the beach and make sure Honu stays on dry sand sections of the beach
P: Honu could possibly hit people
S: IR sensors with a 60 ms reactions time are added to Honu
P: The saline properties near the ocean could corrode or cause Honu to rot
S: The wood is sealed with a marine wood finish, the same finish used on wooden boats. It was also waxed for further sealing.
In order to improve Honu, a wider material selection could also be considered. Our team was considering changing the marine wood shell to carbon fiber so that it would last longer. Another material considered was acrylic. Both of these options were ruled out due to material costs, however if extra funding were obtained it becomes more possible. The treads could also be replaced by rubber instead of wood. Again, wood was chosen more for an expense constraint.
Other considerations that had yet to be explored was the accuracy of the route planning algorithm of Honu. It works out in coding, but it would need more fine tuning and tweaking in order to be considered fully developed. More sensors and more planning would also need to be done in order to collect trash closer to the water as well.
Testing of the GPS accuracy and preparation in case it loses signal would also need to be explored more. There are fail safes. Honu would stay still when it loses signal, but it would be better to explore more on how doing this would affect other components as well if at all.