Introduction

1. Revision history: This is the first revision!
2. Introduction: The ball shooter finds targets and fires balls into them
3. Purpose: This subsystem addresses the ball shooting task for RobotX
4. Scope: This document describes the ball shooter at a conceptual level. Research is limited to the configuration level. Design details are limited to subsystem descriptions and requirements.

The Design

Requirements

1. Core Functionality: The ball shooter is a self-contained subsystem of the WAM-V. It includes subsystems for ranging, fire control, and firing of racquetballs.
   1. The ball shooter shall locate targets
   2. The ball shooter shall shoot balls
   3. The ball shooter shall be capable of firing multiple shots
   4. The ball shooter shall be splash-proof
2. Performance Metrics:
   1. Number of shots (4 balls)
   2. Shot dispersion (CEP 0.2 m) (CEP should be at a specific range)
   3. Muzzle velocity (??? m/s) (should be driven by max range)
   4. Target localization (??? ???)
   5. System weight (10 kg)
   6. System footprint (0.3 x 0.3 m)
   7. System safety (very)

Research / Alternative Enumeration (and analysis)

1. Ball shooting This is commonly accomplished in one of three ways: flywheel, slingshot, or pressure. Flywheel shooters store energy in spinning flywheels which accelerate the ball as it passes by them. Slingshot shooters store energy in elastic bands and which is released to launch the ball. Pressure shooters store energy in pressurized gas or combustibles and use this pressure to accelerate the ball through a barrel. Other methods include punchers, and _____/

   1. Slingshot (or elastic) Pros: mechanical simplicity, easily repaired Cons: elastic fatigue (reproducibility), low speeds

      

      Catapult style “slingshot” shooter

   2. Flywheel Pros: easy reload, compact, good speeds, somewhat reproducible Cons: difficult to model, heavy

      

      Left: two-wheel flywheel shooter for a robotics competition Right: jugs machine used in sportsball training

   3. Pressurized Pros: high speeds, easily modeled, highly reproducible, mechanical simplicity Cons: pressure safety, less compact

   

   Potato cannon design with manually actuated valve

2. Aiming Only real alternatives are to rigidly fix the shooter to the WAM-V or to independently aim it. Rigid aiming requires moving the entire vehicle to reach different aim points. Independent aiming introduces significant mechanical, electrical, and control complexity … but … should allow for much more accurate shooting. Aiming/pointing methods should be investigated in a separate design review and treated as a separate subsystem.

3. Ranging and fire control These are methods for localizing a target, finding a firing solution to hit the target, and actually firing a ball. The fire control problem is effectively “solved” given accurate range, heading, and environmental inputs. Fire control algorithms should be investigated in a separate design review and treated as a separate subsystem.

Ranging / target localization methods should still be investigated. Alternatives include offboard localization (leverage the WAM-V semantic map), onboard classical CV localization, and onboard NN localization.

Own ship dynamics are also important to accurate shooting. Estimates of speed, heading, roll, pitch, and more dynamic variables must be made. The three alternatives are offboard estimation, onboard estimation, and no estimation.

1. Offboard localization
Pros: significant reduction in compute requirements, benefit from parallel development, no additional cameras
Cons: require accurate transforms, tight coupling with WAM-V
2. Onboard classical CV localization
Pros: loose coupling with WAM-V, leverage knowledge of task
Cons: increased compute requirements, additional cameras, susceptible to lighting conditions
3. Onboard NN localization
Pros: loose coupling with WAM-V, robust to lighting conditions, configurable targeting
Cons: Poor NN range-finding, increased compute requirements, additional cameras
4. Offboard kinematics estimation
Pros: reduced compute requirements
Cons: require accurate transforms, tight coupling with WAM-V
5. Onboard kinematics estimation
Pros: loose coupling with WAM-V
Cons: additional sensors, increased compute requirements
6. No kinematics estimation
Pros: very easy
Cons: less accurate (how much?)

Key Assumptions

1. Firing standard racquetballs (Penn Ballistic 2.0)
2. No combustibles
3. Power provided by WAM-V

Design Details (selected design)

1. Description The proposed conceptual design separates the shooter into three subsystems: ball shooting, aiming, ranging & fire control.
   1. The shooting subsystem will be a pressurized air launcher. This is driven by the desire for high accuracy at (relatively) long ranges in the ball shooting task. Pressurized air allows for highly reproducible shots at very high speeds. This was one of two methods seen on other vehicles at RobotX 2024.
   2. The aiming subsystem will be an independently aimed pan/tilt system. While independent aiming adds complexity, it also allows for much high accuracies. With a rigidly fixed launcher, accuracy is dependent on extremely precise control of the overall vehicle.
   3. Ranging & fire control will use onboard target localization and a mix of on/offboard kinematic estimation. Ideally an onboard OAK-D which yaws but does not pitch with the pan-tilt performs target identification. An onboard IMU will estimate pitch and roll (and their rates). Speed and heading estimates will be provided over a data link. An onboard camera and IMU allow for loosely coupled development. The choice of an OAK-D allows for a combined classical CV and NN approach where the strengths of each can be leveraged.