ECE 110/120 Honors Lab Section : Jumping Pixar Lamp

Video: https://drive.google.com/file/d/17chB-rC786Jo12h0CWso9_LG5X9iD6bV/view?usp=sharing 

Report: https://docs.google.com/document/d/1PL7kF8-ZAWcMeF-PUJoHj034pNAhWtdXj_NcEEpQJA0/edit#

Statement of Purpose

The purpose of this project is to make a jumping robot that mimics the lamp from Pixar animation intro [1]. We want the lamp to be able to make a jump at least every 2 seconds and be able to jump half its height, so it would be able to move forward at a reasonable speed and hop in place to squash something short. There is going to be a white LED that turns on when someone waves their hand in front of it, making it an actual lamp. We also want to make it possible to design a shell around the robot, so that it actually looks like the Pixar lamp (if there is time). This project could be useful for entertainment, promotional material, and as inspiration to people of all ages bridging animation and robotics.  

Background Research

We looked at many similar robotic projects, but there doesn't seem to be one combining the Luxo Jr. look with an agile jumper:

• Realistic Pixar Lamp:
  • Dheera created a very realistic-looking robot that uses servos to control each of its joints and a bright NeoPixel ring as the light source [2].
    
    It features a bright light and very cute motions, but doesn't hop very far. We learned from this project that white 3D print filament looks very much like material in the animation, and that some servos are enough to move an entire robot of this size. The jumping style of launching the head then pulling the body along works, but we want to make the jumping faster and more authentic to the animation. 
    
    

• Direct-Drive Hoppers

  • Direct drive hoppers have rigid legs connected directly to a high power density (lots of power for its weight) actuator, usually an electric motor. One example is the Penn Jerboa, shown on the right in the image below. It rotates 2 legs just like the animal jerboa to propel itself forward and waves around a long tail to balance between hops. These robots can jump very quickly and precisely, but can't jump very high. This is because almost all of the kinetic energy is dissipated on impact with the ground instead of stored for the next jump. Also, high power density motors tend to be large, so there are few small direct drive robots [3].
    
    We learned from these projects that in order to jump high, it is beneficial to store energy from the landing impact somehow, like using springs or elastic material. Tails may also be useful to help balance during the jumps, though we'll want to shorten it if we use one so it does not ruin the lamp look. Also, we probably won't be going with a type of direct drive hopper because much of the effectiveness relies on the quality of the motor, and the size is probably beyond the scope of this project.
    
    
• Parallel-Elastic Hoppers
  • Parallel-Elastic refers to the actuator directly connected to the hopping foot but there is an elastic element also connecting the rest of the robot to the foot, so that upon landing energy is stored in the spring. This means that the robot acts like a pogo stick, and is able to release that stored potential energy on the next jump, which reduces the power that the motor needs to supply. The actuator and elastic in parallel means that the motor cannot retract without fighting the spring, but we are able to control the position of the foot exactly and charge up the elastic even when in the air [4].
    
    

  • The Disney Research team created a hopping robot that uses a voice coil actuator which extends and retracts a foot. There are 2 servos controlling the angle of the leg and an IMU to detect the robot angle, so it is able to autonomously jump in place [5].

    
    The voice coil takes a lot of energy, it seems from all of the battery packs on top. We probably won't use an actuator like that, but we will learn from this to use an IMU and some sort of mechanism to keep itself balanced during hops.

  • Another method to jump is to use a cam to wind up a spring. This makes it so that the charging process is over a longer period of time than the energy release, so the power output of the motor can be decreased and so the robot can be much lighter. This particular design from EFPL is inspired from a grasshopper, has a parallel 4-bar linkage to drive the foot, weighs only 7 grams, and can jump 1.4m high, higher than most robots [6].

    
    The cam method seems very promising to us because we are able to provide lots of jumping power with weaker motors. As long as the cam doesn't take too long to wind up, it could work for the robot lamp. In addition, the parallel linkage keeping the body parallel to the ground seems like a good way to make jumps consistent.

  • Bow-legged jumpers have legs made out of an elastic material so there's no need to add other elastics such as spring steel or rubber bands. This design uses a motor to wind up a string to bend the leg inwards, storing potential energy for the jump. The string is quickly released and the bow extends to push the robot up [7].

    

    This is simple but can take a long time to wind up between jumps. Since we want the robot lamp to jump continuously, we probably won't use a string wind up system.

• Series-Elastic: Salto and Salto-1P are ~100g robots built by Berkeley that tries to maximize a metric they call the vertical jumping ability. This is a measure of how quickly the robot can continuously jump upwards, in other words how many meters per second an elevator would have to be traveling in order to catch up with the robot scrambling up tree branches at full speed [4]. So, this metric rewards those that can jump high with each jump but also have little time between jumps. Salto is based on the galago, which has the highest vertical jumping ability of the animals. Research on the physiology of galagos found that they have a mechanism that changes the mechanical advantage depending on the stage of the jump. When the galago is landing one jump, the mechanical advantage is low, meaning that the ratio of the muscle/tendon force to the foot force is low, meaning that a lot of the kinetic energy of impact can be stored back into the elastic tendons over a long period of time. After that, when the galago is ready to spring back up into the next jump, the mechanical advantage of the muscle to the foot suddenly increases, meaning that the force from the muscle and tendons are amplified many times and released in a quick burst faster than the "charging" motion [8]. So, the galago is able to release more energy in the upward motion than otherwise possible with the muscle and tendons alone. 

               
  The Salto robot has a vertical jumping ability of 1.75m/s, the highest recorded of any robot (Galago is 2.25m/s and the second highest was a 4-legged direct-drive jumper called Minotaur) [4]. We will base the jumping mechanism on our hopping lamp on Salto's, because it is so effective and contains relatively few components. Salto uses carbon fiber rods for the leg linkage, a brushless DC motor (Scorpion S-1804-1650KV) and surgical tubing torsion spring for the main leg, a smaller motor for spinning the flywheel tail, a 11.1V LiPo pack, and a small control board designed specifically for flying robots [4]. 

  Overall, the software is more complex because the robot has to precisely control its angle in the air during jumps. Salto-1P, the second version, needed two propellor thrusters so it could balance side to side [9]. However, their code is throughly described in [10] and is on Github, so we think it wouldn't be too hard to get a simplistic version of it working. 

We also looked at the mechanics of running and hopping:

• This paper analyzes legged bouncing animals, and finds that animals with a longer time spent in stance (crouched down position) can hop higher [11]. This is relevant to how many of the jumping robots created have been biomimetics, or robots mimicking a certain animal such as galago or grasshopper.
  

• There are different types of spring loaded running: sagittal or transverse refers to whether the movement is on the vertical plane or the horizontal plane. Also, the SLIP model stands for Spring Loaded Inverted Pendulum [12]. This video helped us understand the vocabulary and basics of what the technical papers are talking about.

We want to make a robot that hops well and could resemble the Pixar lamp because no one seemed to have done this before and it would be a great way to learn about small electronics, mechanics, and control theory.

Block Diagram / Flow Chart

System Overview

Thick arrows represent power, and thin lines represent data.

• The entire robot is powered by a LiPo pack, which is converted to 5V via a step down converter for the Teensy control board.
• The LED and photodiode manage the lamp feature. If the photodiode detects a sudden absence of light then back to "normal" light, then someone probably waved their hand.
• The IMU and rotary encoder on the leg feed information to the Teensy so it can balance and plan leg extensions with code.
• A weaker brushless motor spins a flywheel like in the Salto project to balance the robot during jumps.
• A stronger brushless motor is connected to a gearbox, then to a spring, then to a parallel linkage that extends or retracts the hopping leg.

We may swap out one of the motors with a servo, or brushed DC motor. Initially we are looking to make the robot hop in place autonomously, but will add remote control later.

Parts

Possible Challenges

1. Salto uses thin carbon fiber rods. We don't plan to use carbon fiber because it is annoying to cut safely and is expensive. We are using either wood or aluminum tubes, so our robot leg may be too heavy to jump as well.
2. Like Salto, we are going to need complex software to control jumps in order to land in a way that allows the next jump within a short amount of time, with information from both the leg encoder(s) and IMU.
3. Neither of us have worked with such lightweight motors and electronics before, so we need to do a lot of research on which to choose.

References

[1] Ferangelini. 2021. Pixar lamp intro from pixar movies HD 720p. [online] Available at: <https://www.youtube.com/watch?v=PGKmexNTHNE> [Accessed 19 September 2021].

[2] Dheera. 2021. GitHub - dheera/robot-luxo. [online] Available at: <https://github.com/dheera/robot-luxo> [Accessed 19 September 2021].

[3] Kenneally, G., De, A. and Koditschek, D., 2016. Design Principles for a Family of Direct-Drive Legged Robots. IEEE Robotics and Automation Letters, 1(2), pp.900-907.

[4] Haldane, D., Plecnik, M., Yim, J., and Fearing, R., 2016. Robotic vertical jumping agility via series-elastic power modulation. Science Robotics, 1(1), pp. eaag2048, doi: 10.1126/scirobotics.aag2048.

[5] DisneyResearchHub. 2021. Untethered One-Legged Hopping in 3D Using Linear Elastic Actuator in Parallel (LEAP). [online] Available at: <https://www.youtube.com/watch?v=M0ZXmGRCuts> [Accessed 19 September 2021].

[6] M. Kovac, M. Fuchs, A. Guignard, J. Zufferey and D. Floreano. "A miniature 7g jumping robot," 2008 IEEE International Conference on Robotics and Automation, 2008, pp. 373-378, doi: 10.1109/ROBOT.2008.4543236.

[7] G. Zeglin and B. Brown, "Control of a bow leg hopping robot," Proceedings. 1998 IEEE International Conference on Robotics and Automation (Cat. No.98CH36146), 1998, pp. 793-798 vol.1, doi: 10.1109/ROBOT.1998.677082.

[8] Aerts P. (1998). Vertical jumping in Galago senegalensis: the quest for an obligate mechanical power amplifier. Philosophical Transactions of the Royal Society B: Biological Sciences, 353(1375), 1607–1620. https://doi.org/10.1098/rstb.1998.0313

[9] Haldane, D. (2017). Rapid and Agile Locomotion with Power-dense Millirobots. UC Berkeley. ProQuest ID: Haldane_berkeley_0028E_16846. Merritt ID: ark:/13030/m5wx2ckp. Retrieved from https://escholarship.org/uc/item/9xj2099q

[10] D. W. Haldane, J. K. Yim and R. S. Fearing, "Repetitive extreme-acceleration (14-g) spatial jumping with Salto-1P," 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017, pp. 3345-3351, doi: 10.1109/IROS.2017.8206172.

[11] Blickhan, R. and Full, R. 1993. Similarity in multilegged locomotion: Bouncing like a monopode. Journal of Comparative Physiology A, 173(5), pp. 509-517, doi: 10.1007/BF00197760.

[12] Koditschek, D. 2016. Running like a spring-loaded pendulum. University of Pennsylvania. [online] Available at: <https://www.coursera.org/lecture/robotics-mobility/2-1-2-running-like-a-spring-loaded-pendulum-16uwN> [Accessed 7 September 2021].