ETS-VII

Terrestrelasso, via Wikimedia Commons. CC BY-SA 4.0.

Overview

Engineering Test Satellite VII was a Japanese two-spacecraft mission demonstrating autonomous rendezvous and docking and free-flying manipulation. It launched on 28 November 1997 [2]. The program is described in full by its project lead in [4]. The chaser, Hikoboshi, carried a 2 m six-degree-of-freedom manipulator; the smaller target, Orihime, was released and recaptured [2].

Specifications

Parameter	Value	Source
Chaser mass	2550 kg
Chaser inertia, Ixx / Iyy / Izz	6200 / 3540 / 7090 kg m2
Target mass	500 kg
Manipulator	2 m, 6 DOF
Arm mass	about 140 kg over six links	[1], [5]
Arm attachment point in base frame	(-0.79, -0.29, 1.00) m

Rows with no marker are from [1].

Mission profile

Parameter	Value	Source
Launch	28 November 1997, H-II Launch Vehicle 6F from Tanegashima Space Center	[2], [11]
International designation	Hikoboshi 1997-074B, Orihime 1997-074E	[11]
Combined launch weight	2,860 kg	[11]
Orbit	circular, 550 km altitude, 35 degrees inclination, period about 96 minutes	[1], [11]
Design life	1.5 years	[11]
Rendezvous-docking experiments conducted	3	[11]
Ground link	NASDA ground stations via the TDRS data relay satellite	[11]
Planned end of the experiment program	end of May 1999	[1]
Extended to	December 1999	[1]

Manipulator

The arm masses about 140 kg across six links and carries a multisensory gripper and an advanced robotic hand supplied by MITI [1], [5]. Link parameters, which are what make the free-floating dynamics computable, are published in full [1]:

Link	Length (m)	Mass (kg)	Izz (kg m2)
1	0.35 (offset 0.28)	35.0	1.69
2	0.87	22.5	3.75
3	0.63	21.9	2.53
4	not stated	16.5	0.072
5	not stated	26.0	0.13
6 (with end effector)	0.53	18.5	0.26

CCD cameras are carried at the shoulder and on the hand, and the hand carries force-torque sensing used for compliance and impedance control during contact tasks [6].

Free-floating dynamics

The arm masses about 140 kg on a 2550 kg base [1], so arm motion produces measurable base attitude change and the attitude controller’s response feeds back into the arm. The two systems are coupled through the base and neither can be designed in isolation.

The Generalized Jacobian Matrix maps joint rates to inertial-frame hand motion including that reaction, allowing resolved motion-rate control in inertial space on a free-floating base [1]. In flight, a 200 mm straight-line inertial path tracked at 10 mm/s held hand orientation at -22.8 degrees in the inertial frame while the base pitched 0.6 degrees under the reaction moment.

Reaction Null-Space control finds arm motions that produce zero net reaction on the base [5], and the same formulation applies to a manipulator on a flexible base, where the target is suppression of structural vibration rather than of rigid-body attitude change [9]. With a six degree-of-freedom non-redundant arm the null space available for this is n minus 3, so three of the six joint freedoms remain for reactionless motion [1]. Attitude variation during the final approach phase of a capture was held to 0.5 degrees peak to peak.

Gravity gradient torque is not negligible on this scale. In free drift the base attitude drifts more than one degree over several minutes [1], and beyond one degree of pointing error the high-bandwidth link is lost, which bounds how long the vehicle can be left in free-floating mode during an arm operation.

Teleoperation under delay

Round-trip delay from the NASDA ground station through TDRS was 6 to 7 seconds [3]. Motion commands were uploaded at 4 Hz in an isochronous synchronized stream [1].

Two compensation approaches were flown. The DLR GETEX experiment used task-directed programming with a predictive three-dimensional graphics display, an approach inherited from ROTEX, in which the operator commands a simulated arm on the ground and the flight arm follows [8]. The Kyoto and NASDA experiment instead closed a direct bilateral coupling with kinesthetic force feedback across the 6 to 7 second delay [3], which is the harder case because the delay sits inside the force loop rather than outside it. Under that scheme a slope-tracing task and an 18 mm peg-in-hole insertion with 0.4 mm clearance were performed without visual feedback, at a hand speed limited to 2.0 mm/s for safety.

Rendezvous and docking

Sensing is staged by range [2]:

Phase	Sensor	Range
Relative approach	GPS relative navigation	500 m and beyond
Final approach	rendezvous laser radar	500 m to 2 m
Docking	proximity sensor (PXS)	2 m and closer

The vehicle holds at a 2 m virtual point for 15 minutes before the docking phase [2]. The handover between sensors, rather than any one sensor’s accuracy, is the design problem: each stage must acquire before the previous one loses the target, and the mission overview identifies autonomous rendezvous and docking as one of the two technology sets the mission existed to prove, alongside free-flying manipulation [4]. Docking approach velocity is 1 cm/s, contact acceleration peaks at 14 mm/s2, and the low impact specification is below 1.5 mg [2]. Achieved docking position accuracy under remote-piloted control was 1 cm.

Three flight programs were run. FP1 on 7 July 1998 separated at 1.8 cm/s and redocked successfully [2]. FP2 began on 7 August 1998, reached 12 km of separation, encountered a thruster pulse injection failure at intervals of one to a few seconds during final approach, and docked successfully after recovery on 27 August 1998. FP3 ran on 26 and 27 October 1999 and included an R-bar approach initiated 9 km from the target, with remote-piloted sequences flown from 2 m out to hold points at 6 m and 12 m [2].

Manipulation experiments

Four experiment categories were flown: ground teleoperation under large time delay, orbital replacement unit exchange, space truss structure deployment and assembly, and antenna assembly [4]. Antenna reflector assembly was performed by teleoperation with vision and task-level sequencing [7]. The advanced robotic hand was evaluated separately against the ORU exchange and truss handling tasks, with compliance and impedance control closing on its force-torque sensor [6]. The GETEX sequence added a task-directed programming layer over the same hardware [8].

Capture of the free-flying target was performed with the hand starting about 400 mm from the grapple fixture against a target drifting at 10 mm/s, with the controller switching mode when hand position error fell below 100 mm [1].

Ground segment

Operations ran from the NASDA ground station through TDRS [3]. Three successive flight passes were used for the experiment set of 30 September 1999 [1], which is the shape of the constraint: work is bounded by pass length, not by arm speed. A summary of the mission’s dynamics results and their relation to the laboratory work that preceded them is given in [10].