Introduction

1. Revision history: This revision builds on learnings from a proof of concept model and (somewhat) formally establishes the needs for this project.
2. Introduction: The sensor turret mounts high-quality and low-FOV sensors to a stabilized and aimable platform attached to the blueboat.
3. Purpose: The sensor turret serves three purposes: providing steering-independent sensor data, providing blind-spot coverage, and adding high-quality sensors to the blueboat sensor suite.
4. Scope: This document describes the need for the sensor turret and several CONOPs. It reviews common designs and describes a proof-of-concept design. From this review, it establishes requirements for the sensor turret. Finally, the document describes potential concepts and the next steps for the project.

The Design

Requirements

1. Core Functionality:

   1. The system mounts to the blueboat
   2. The system carries “high-quality and low-FOV” sensors including
      1. LIVOX AVIA
      2. CS-mount RGB camera
   3. The system stabilizes these sensors relative to the horizon
   4. The system allows for pointing of these sensors in (at least) the yaw axis

   The combination of these core functions enables the following operation modes:

   1. relative region scanning: The turret repeatedly sweeps a convex set of angles relative to the boat. This provides blind-spot coverage functionality. This mode also allows for setting a fixed relative heading.

   i.e. the turret scans the 180 degree region to the port of the boat once every 2 seconds
   2. global region scanning: The turret repeatedly sweeps a convex set of angles which bound a region of interest on the map. This provides a steering-independent perception functionality. This mode also allows for object tracking.

   i.e. the turret scans the entire “follow the path” task as the boat circles it i.e. the turret watches a changing light pattern while the boat stationkeeps

2. Performance Metrics:

   1. Tracking Performance

      1. Scanning envelope: what are the ranges of angles covered by the aimable system?

         1. Yaw should allow for any blind spot coverage (i.e. 360 deg). Allowing for >360 degrees would require a slip ring and might be overly complex … yaw +/- 180 deg

      2. Aiming accuracy: how near is the pointing angle to the commanded angle?

         1. The LIVOX FOV is (either) a 70 deg cone or a 70x4.5 deg rectangle. The CS-mount camera FOV depends on lens, but should be of a similar magnitude. As long as the true angle is known … effective sensor use only requires that the object/angle of interest is within the field of view. roll/pitch +/- 2 deg yaw +/- 1 deg (this adds the requirement that the true angle relative to the blueboat is known to a high degree of accuracy (<0.1 deg))

      3. Scanning frequency: what is the maximum scanning frequency for a given region?

         1. The robotx competition does not involve dynamic maps, so (in theory) a single scan can capture all necessary information. This requirement should be set after discussion with the software team.

      4. Scanning speed: what is the maximum angular rate of change?

         1. This requirement is driven by both the scanning frequency requirement AND by the necessary speed to track an object of interest. This requirement should be set based on the expected operating conditions (as outlined after discussion with the software team)

         For reference: the maximum rated speed for the blueboat is 3 m/s. For a straight path with a 1 m nearest approach, the maximum angular speed should be 170 deg/s (this is very high)

   2. Stabilization performance
      1. Stabilization envelope: what is the maximum angular offset which can be counteracted
         1. Roll and pitch should account for tilt during acceleration and due to moderate waves roll/pitch +/- 30 deg
      2. Frequency Response: what frequencies and amplitude disturbances can be mitigated?
         1. This requirement is driven by wave disturbances. Tilt from acceleration can be treated as a zero-frequency disturbance with a max amplitude described by the stabilization envelope. From experience/vibes , bad waves are a reasonable worst-case estimate and emulate roll/pitch disturbance of frequency 1 Hz and amplitude 30 deg
      3. Stabilized payload: what is the maximum mass for which the system meets other stabilization requirements?
         1. The system should stabilize the LIVOX AVIA (\~500g), a CS-mount camera (\~50g), and lens (~50g). Some allowance should be made for mounting hardware and additional sensors. max payload 1.0 kg
   3. Physical compatibility
      1. blueboat compatibility: can the turret be mounted and carried by the blueboat?
         1. The blueboat max payload is 15kg. asdl-blueboat-ext adds ~6kg of payload. max turret (with payload) mass 3 kg
         2. Existing sensors and antennas restrict view. turret FOV should not be significantly obstructed
         3. Forward mounting points on asdl-blueboat-ext are largely occupied. The aft crossbar is used to mount the e-stop and light tower. Remaining mounting space is to the aft port/stbd hull. turret footprint should fall within blueboat footprint
         4. should consider ease of removal for transport (expensive sensors!!!)
      2. sensor compatibility
         1. The aimed volume should account for the planned sensors
   4. Electronic compatibility
      1. The payload sensors are PoE compatible, the turret need not interface with the sensors.
      2. Voltage is available over PoE, at 56V, 13.8V, 5V, and BAT voltage. The turret electronics should use available voltages.
   5. Software compatibility
      1. A ROS2 node should expose actions associated with the proposed operation modes

Research

1. Alternative Enumeration: how is this done today?
   1. mechanical alternatives
      1. number of axes: stabilizing gimbals typically use 2 or 3 orthogonal axes
         1. 2-axis: can EITHER stabilize pitch and roll at fixed heading OR provide pan-tilt pointing without roll stabilization.
         2. 3-axis: can provide BOTH pan-tilt pointing AND pitch-roll stabilization
      2. axis arrangement: 3-axis stabilizing gimbals have different axis arrangements. 2-axis gimbals use only orthogonal axes.
         1. orthogonal axes: 3 axes are orthogonal to each other
         2. 45 deg offset: middle axis is 45 degrees offset to first and last axes
      3. motor choice: type of motor to use
         1. geared motors (brushed or brushless): high gear ratio motors
         2. direct drive motors (brushless): use field-oriented-control for position servoing
   2. controller alternatives
      1. control board: high-frequency controllers which drives the stabilization loop
         1. off-the-shelf brushless-gimbal-controller (BGC): commercially produced control boards
         2. custom controllers: raspi or microcontroller with custom software to interface with custom motor controllers
      2. sensors: information available to controller
         1. payload IMU: accelerometer and gyro on payload estimate angle relative to gravity. motor feedback uses coupled signal
         2. payload+board IMU: accelerometer and gyro added to base to better estimate angle relative to gravity
         3. IMUs + encoders: add on-axis encoders to improve motor feedback
2. Alternative Analysis: what are the pros+cons of these methods
   1. mechanical alternatives
      1. number of axes: 3 axes should be used
         1. improved stabilization: stabilization is a key benefit of the turret system. A 2-axis turret can only provide stabilization if the oscillation axis is aligned with the pitch axis. This means that a 3-axis system is required for stabilization on all headings EXCEPT the two which align with waves.
      2. axis arrangement: the 45 deg offset should be used
         1. reduced volume: In order to guarantee passive stability, the payload CG should be underneath roll/pitch axes. By angling the middle axis (likely roll) up, the motor and pivot hardware can be placed beneath the payload and reduce the cross-section of the turret. (also it looks cool …)
         2. easier (?) balancing: the 45 deg middle axis offset means the “back” of the payload is exposed. Normally this is useful to see camera screens, but it also provides volume for ballasting.
         3. might even be worth consider the (abnormal) configuration with angle roll AND pitch axes
      3. motor choice: depends on the control board … but lean towards direct drive BLDC
   2. controller alternatives

Key Assumptions

1. why those decisions

Design Details (selected design)

1. Description: pictures, sketches, block diagram, CAD, algorithm, …
2. Interfaces: what data/resources enter and exit this system
3. Performance Analysis: estimation of performance metrics (math preferred)
4. Risk & Mitigation: how can this design fail & backup plans (redundancy + what is hard?)