Authors: Robert Mahony, Jonathan Kelly, Stephan Weiss
Organizations: Systems Theory and Robotics Group, Australian National University, Australia · Space and Terrestrial Autonomous Robotics Systems Laboratory, University of Toronto Canada. · Control of Networked Systems group, University of Klagenfurt, Austria.
In this paper, we present a robotics-tailored exposition of Galilean symmetry that leverages the community's familiarity with and understanding of rigid-body transformations and pose representations. Our approach contrasts with common treatments in the physics literature that introduce Galilean symmetry as a stepping stone to Einstein's relativity. We show that the classical homogeneous matrix algebra known well to roboticist translates directly into the Galilean modeling framework and provides a powerful tool for tackling robotics problems involving inertial measurements and spatio-temporal variables. We demonstrate the potential of the approach by developing a powerful model for studying motion of vehicles in an Earth fixed rotating reference frame, and developing a fusion algorithm for measurements with uncertain time-stamps.
Figures & tables
Figure 1: Consider a point p (black dot) with respect to a moving frame reference frame {A} (red axes). The point is moving with linear velocity vp with respect to a third inertial observer. The frame is moving with linear velocity w and rotating with angular velocity Ω with respect to the same inertial observer. The relative inertial velocity v=(vp−w) of the point is that induced only by velocity of the point and the translation of the frame. The coordinate velocity p˙ of the point is the sum of the relative inertial velocity and the perceived coordinate velocity induced by rotation of the frame.
Figure 2: Galilean transformation ApA↦ApA′ and AvA↦AvA′ . The point ApA (blue) is rotated by Q (dashed orange), translated by q (orange), translated by AtAb (orange) to become ApA′ (green). The velocity AvA is rotated by Q (dashed orange) then boosted by b (orange) to become AvA′ (green).
Figure 3: Spatio-temporal coupling of timestamp and velocity uncertainty under Galilean transformation. Homogeneous uncertainty μpt in the event pt (shown by the dashed covariance ellipse on the left) is smeared by the Galilean transformation G in the boost direction (shown by the expanded dashed covariance ellipse on the right). The figure also shows transformation of the velocity v to Gv with the addition of the boost.
Figure 4: Representation of a concentrated Gaussian on Gal(3) . The Gaussian distribution is defined in Euclidean coordinates ϵ∈Rn on the tangent space TF^G of the Lie-group and then mapped onto a distribution on the group itself using group multiplication, the wedge map ( 11 ), and the exponential map. The Lie-group Gal(3) is drawn as a sphere for simplicity of the diagram and the reader should be aware that the geometry of Gal(3) is more complex than shown here.
Figure 5: Spatio-Temporal stochastic correlation. Here the (7:9,10) and (10,7:9) entries of Q∈R10×10 are non-zero and lead to the covariance correlation Q(p,t) shown between the position p and the time t variables. If time t is measured (blue point on the time-axis) then the uncertainty in position p^ is the conditional distribution shown in the heavy blue line (indicating a collapsed ellipse) on the p -axis. If position p is known (green point on the position-axis) then the uncertainty in time t^ is the conditional distribution shown in the heavy green line on the time axis.
Figure 6: Kinematics of a Galilean Frame. The reference {0} is fixed in time, while frame {A} is the associated stationary inertial frame. The Galilean frame {B} is moving in space-time. The kinematics of the Galilean frame coordinates F=0FBt0 with respect to the reference are given by ( 20 ).
Figure 7: Isochronous Inertial Kinematics of a Galilean Frame. An inertial reference {A} moves with respect to the origin of {0} . The moving target Galilean frame {B} is isochronous with respect to the inertial frame {A} . The kinematics of the isochronous Galilean frame AtFBtAt is given by ( 27 ).
Figure 8: Isochronous Galilean kinematics. Both the reference {A} and target {B} Galilean frames are moving with respect to {0} . The kinematics of the isochronous Galilean frame AtFBtAt is given by ( 31 ).
Figure 9: Galilean kinematics of an agent with respect to a moving frame attached to the surface of the Earth.
Figure 10: Force balance describing the Earth geoid. The geoid is defined by the balance between gravitational acceleration gA and normal acceleration aAn perpendicular to the surface of the geoid to generate the centripetal acceleration aAc=gA+aAn required to keep a point on the surface of the geoid circling the rotational axis of the earth. The proper acceleration aAp=aAn measured by the IMU is the normal acceleration. The perceived gravitational acceleration is g0e3=−aAn .
Figure 11: The measurement yτi is associated with frame {ti} with Galilean coordinates Fti with respect to a reference {A} . The goal is to filter for the state Fk associated with the frame {tk} . The relationship between {ti} and {tk} is given by the Galilean frame tiFtkti=Fti−1Ftk that can be computed by pre-integrating IMU data to compute Ft over the time interval t∈[ti,tk] . Each small hop corresponds to an IMU measurement update tiFtjti↦ΓjtiFtjtiΥj ( 47 ).
Figure 12: Fusion using the Galilean construction (top figure) compared to using classical fusion (bottom figure) for the case where there is high uncertainty in the measurement timestamp. The prior state estimate is the small red circle (at (0,0) ) with red homogeneous uncertainty ellipse (large red circle centred at (0,0) ). The system is travelling at velocity 1m/s as shown by the arrow. The measurement is shown as a cyan cross with cyan homogeneous uncertainty ellipse. The true state is shown by the black asterisk. The fused estimate is the blue circle with blue uncertainty ellipse. Note that for the Galilean fusion the uncertainty in the temporal coordinate stretches the uncertainty of the state estimate in the direction of the velocity. The temporal uncertainty is not modelled in classical fusion and the posteriori state estimate is homogeneous and overconfident.
The special Galilean group, usually denoted SGal(3), is a 10-dimensional Lie group whose important subgroups include the special orthogonal group, the special Euclidean group, and the group of extended poses. We briefly describe SGal(3) and its Lie algebra and show how the group structure supports a unified representation of uncertainty in space and time. Our aim is to highlight the potential usefulness of this group for several robotics problems.
Jonathan Kelly
Space & Terrestrial Autonomous Robotic Systems (STARS) Laboratory, University of Toronto Institute for Aerospace Studies (UTIAS), Toronto, Canada
Multi-sensor state estimation requires accurate spatial and temporal calibration between heterogeneous sensors. Existing spatiotemporal calibration methods typically represent sensor extrinsic parameters as a rigid-body transformation together with a scalar temporal offset, causing the spatial and temporal parameters to enter the estimation problem through separate mechanisms. While the spatial transformation composes directly with the robot pose, accounting for the temporal offset requires evaluating the trajectory at a shifted time, typically through motion compensation, interpolation, or a continuous-time trajectory representation. In this work, we formulate joint trajectory estimation and spatiotemporal calibration using the special Galilean group for the robot trajectory and the Aristotelian subgroup for sensor extrinsics. This representation incorporates the coupling between motion and temporal offset directly through the group structure, avoiding the need for a separately parameterized continuous-time trajectory. We develop twist- and IMU-driven motion priors, accommodate both ideal and non-ideal clocks, and evaluate the framework with camera, position, and range measurements. Simulated and real experiments demonstrate calibration and trajectory-estimation performance comparable to established methods across heterogeneous sensor configurations. Our simplified framework therefore offers a quality-of-life improvement over existing methods without compromising performance.
Abhishek Goudar, Jonathan Kelly, Timothy D. Barfoot
University of Toronto Robotics Institute, University of Toronto, Toronto, Ontario, Canada
Robots exhibit a rich variety of symmetries arising from their mechanical structure and the properties of their tasks. Although many robotics problems exhibit several symmetries simultaneously, existing approaches typically treat them in isolation, failing to exploit their combined potential. This paper introduces cross-space symmetry compositions, a framework for learning robot policies that are jointly equivariant to multiple symmetries across configuration and task spaces. Leveraging the differential-geometric structure of the forward kinematics map, we both descend symmetries from configuration to task space and lift symmetries from task to configuration space, enabling their composition within a unified representation space. We validate our framework on simulated and real-world experiments on a dual-arm robot, demonstrating that jointly leveraging multiple symmetries yields improved generalization. Video and source code are available at https://sites.google.com/view/cross-space-symmetries.