category: SummerBlogs This creates a floating frame with caption below For best results, attach .jpg or .png which has been resized about right Attach:Simon_bdml.jpg Δ | Simon in BDML for spring and summer 2020
Spring/Summer blog - update every few days with new findings.
Just getting settled into the lab. It has been great getting to meet people and to learn about their respective projects. I have started to review previous FlyCroTug designs and results, and to evaluate them against future goals for the project.
On the side, I also got the chance to work with Capella to help with the post processing of some Gecko Adhesive. It was great to get a better feel the manufacturing process and how the Gecko adhesive works!
The rest of the week was spent getting trained for the different spaces and looking into the manufacturing of microspine grippers. Microspine grippers allow the FlyCroTug drones to passively adhere to rough surfaces, such as roof tiles, bricks, or concrete, allowing them to exert much larger forces on the environment. BDML has experimented with several different manufacturing methods to make these grippers. One method uses a combination of resins (one stiff, one elastic) to allow the stiffness of the spines to be tuned in the normal and tangential directions independently. An alternate method sacrifices the tune-ability in favor of weight reduction and manufacturing simplicity, by simply adhering the spines to Kapton film fingers. The latter method seems promising for aerial vehicles, but needs to be recreated and optimized for the future FlyCroTug platforms. I have also been toying with the idea of directly machining microspines and bow-spring style elastic elements from thin sheets of spring steel or Nitinol. This would drastically reduce the manufacturing complexity, but would come at an added weight cost. More tests to come next week!
My second week was full of testing and rapid prototyping! After fixing the lab's Ultimaker 3D printer, it was off to the races, creating new jigs and test assemblies.
Continuing with the Kapton film finger design, I investigated different manufacturing methods to facilitate quickly making new microspine arrays. I tested different adhesives, surface treatments, and jig designs to quickly and accurately align spines prior to bonding them to the adhesion.
The most reliable method I found was as follows:
- Laser cut fingers with the desired size and pitch out of Kapton, fiberglass, or a thin sheet material of choice.
- Rough the ends of each finger with sandpaper and subsequently clean it with isopropyl alcohol.
- Use a 3D printed jig to align the selected fish hooks to the ends of the fingers.
- Apply a generous band of JB weld quick cure epoxy to the fingers and place the jig on top.
- Using a toothpick or similar, brush excess epoxy over the tops of the fishhooks to ensure the center of the hook is fully encapsulated.
- Allow sufficient time for the epoxy to cure to full mechanical strength.
- Carefully remove the 3D printed jig.
- Use side-cutters or similar to cut the shangs from the fish hooks.
- Use an exacto knife or similar to carefully cut through any epoxy bridges between the fingers.
- Enjoy your new microspine array!
Following this testing, I came up with an idea to use ribbons of thin film (such as Kapton or thin fiberglass sheet) to create a cassette of coil springs on which microspines could be attached. This method of manufacturing greatly reduces the complexity, compared to the previous SDM methods, and can be quickly prototyped and tested. This method also allows the normal and tangential stiffness, and the spine travel distance to be easily tuned by changing the width and thickness of each individual "finger". For thin films, "thickness" can be changed by simply layering multiple strips together. Tony helped me further develop the idea and started working on a collapse-able truss design to actuate the gripper. Prototyping hit a bit of a snag due to precautions taken to limit the spread of COVID-19. The labs transitioned to remote work only, and my work on microspines will be put on hold.
Stanford has now officially shut down all "non-essential" on-campus research operations. Students are being asked to leave the campus as a precautionary measure. In response, I have returned home and will work remotely until the social-distancing measures are reduced. It took some adjustment to remote work and to shift the focus of my research away from hardware and towards software. One of the key aspects of the FlyCroTug project where there is the potential for some development is on the automation of the perching procedure. Prior investigations by BDML in perching utilized human pilots with constant line-of-sight on the vehicle. Even under these conditions, not all perch attempts are successful. This is in part due to the complexity of the maneuver and in part due to the stochastic nature of the surface asperities which micro-spine grippers attach themselves to.
Previous work has identified optimal flight trajectories (target normal and tangential velocities) for a perched landing. Additionally, some characterization of perching failures was conducted. This included the identification of failed vs. successful perching attempts using the quadrotor's sensory information. Acceleration data was found to be the most useful and previous experiments showed 90-95% identification accuracy. I have started investigating the state-of-the-art for non-linear quadcopter control. A nonlinear controller will likely be required in order to maximize the change of success on a vertical perching attempt, which requires a 90 degree pitch turn. Automating perching will be necessary when the drone is being flown autonomously, but could still be beneficial when manually flying with limited line of sight or a high-latency video stream. Additionally, automated perching maneuvers could improve perch location precision, allowing for specific vantage points to be used, and reducing the minimum landing site size requirement.
This marks the end of my first week in isolation - fortunately my room has plenty of natural light and I have no shortage of reading. This week has been spent continuing to learn about non-linear control techniques for quadcopters, including methods developed by ETH Zurich's Flying Machine Arena and UPenn's GRASP Lab. In preparation for future simulation work, I have also been brushing up on ROS and Gazebo.