The actual question
Four-wheel omnidirectional robots can translate along two body axes while rotating. Their freedom to move is not unlimited: a requested chassis velocity has to be translated into four separate motor speeds, each with a finite ceiling.
This paper asks how the kinematic map, a smooth movement cost, and the sharp actuator limits fit together. It then uses the configuration space of planar rigid bodies to describe pose error without pretending heading is an unrestricted real number. The work is an idealized analytical and numerical study; the proposed VEX V5 application remains to be validated on hardware.
From chassis twist to wheel rates
The pose belongs to SE(2), the group of planar translations and rotations. Represent the chassis as a homogeneous transform whose rotation is R(θ) and whose position is p = (x,y). The body twist ξ = (vₓ,vᵧ,ω) contains two body-frame velocities and one angular rate.
The symmetric X-drive’s wheel-speed map is φ̇ = (1/r) J ξ, with four rows of J and three independent twist variables. Here r is effective wheel radius, R is effective yaw lever arm, and the motor ordering fixes the signs. The paper calculates rank(J) = 3 and JᵀJ = diag(2, 2, 4R²), establishing that the ideal kinematic map can represent all three planar twist directions.
This is a rank and inversion result, not evidence of arbitrary acceleration or traction. The geometry must be calibrated on the real chassis because wheel placement, signs, contact conditions, and effective radius alter the map.
One metric, two very different boundaries
Squaring and summing the four wheel angular speeds produces a positive-definite quadratic form. That defines a Riemannian metric: an infinitesimal way to measure movement cost. In the paper’s symmetric model it is flat, so an unconstrained, fixed-duration minimum of this cost uses straight world-frame translation and constant angular rate. It is not a measured electrical-energy model.
The hard speed bound is not an ellipse. With each wheel satisfying |φ̇ᵢ| ≤ Ω, the same model yields an exact polyhedral feasible set:
Because the smooth quadratic form and piecewise-linear wheel ceilings encode different questions, a cost-minimizing path is not automatically executable. The paper gives restricted exact travel-time benchmarks for fixed-heading translation and pure rotation, not a general solution for obstacle-filled trajectory planning.
The local SE(2) regulator
Let Td be a stationary target pose and E = T⁻¹Td the relative displacement. Taking the matrix logarithm converts the error locally into three coordinates e = log(E)∨. Near the desired pose, the controller points along this error and scales its command until all wheel limits are met.
That proof depends on the local logarithm chart and avoids the heading branch cut near ±π. A moving target, sample-and-hold motor commands, wheel slip, and estimation noise require separate analysis. The manuscript adds a bounded-disturbance calculation to explain why persistent uncertainty generally creates a residual error neighborhood rather than exact convergence.
What the algebraic equilibrium study establishes
Computational algebra can characterize roots and branches when a subsystem is truly polynomial. The paper uses a parameterized polynomial heading model to show how elimination and root analysis can reveal equilibrium branches. It carefully notes that the real SE(2) logarithm, trigonometric geometry, absolute values, and saturation are not globally polynomial.
Consequently, the toy polynomial example is not evidence of an actual drivetrain bifurcation. Making that claim would require a physically justified model with measured motor and contact dynamics, then validation against experimental data.
Results, interpretation and next steps
Reproducible ideal-model numerical integrations in the manuscript illustrate the stationary-pose regulator’s predicted convergence and enforcement of commanded wheel-speed bounds. Their value is checking a mathematical derivation, not estimating physical reliability.
The proposed implementation pathway is to identify the real Jacobian, calibrate encoder and inertial-sensor signals, estimate pose, log actual wheel motion, test against a conventional baseline, and measure how slip, sampling and disturbances change the result. Until those tests exist, this page describes analytical behavior, not demonstrated autonomous navigation.