# Compensation of asynchronous time of flight measurements in a long baseline navigation

**Part of the following topical collections:**

## Abstract

In a long baseline (LBL) acoustic positioning system, a navigation subject estimates its own position at a time based on the ranges between itself and the LBL transponders. To obtain the ranges, each transponder sends acoustic wave to the subjects receiver. Here, a range equals to time required by the wave to travel between these two points multiplied by the wave propagation speed. This method is known as the time of flight (ToF) measurement. One of the ideals in carrying out a ToF is that the navigation subject would remain still throughout the measurement. Corollary, position estimation based on the ToF holds on the same ideal. However, an acoustic wave propagates in a very low speed. Due to this characteristic, the displacement of a moving navigation subject like an autonomous underwater vehicle (AUV) may not be negligible and becomes a source of bias for the range estimations. The AUV motion would also lead to asynchronous ToFs, i.e. acoustic waves sent by transponders arrive at the AUV in different time epochs. In this paper, we present LBL navigation for an AUV that deals with uncertainty due to motion of the AUV. Here, state estimator is modeled while considering the bias for each ToF. Once we obtain the state space representation, we apply Kalman filter to estimate the AUV position and speed. By simulation, we demonstrate that the estimator follows the actual states with good accordance.

## Keywords

Autonomous underwater vehicle Inference diagram Long baseline Sonar Underwater acoustics## 1 Introduction

Initial development of concepts and applications of underwater acoustic positioning systems can be traced back to 1960s [1, 2]. For underwater applications, these systems are preferable since position references from the more convenient global positioning systems (GPS) are not available in the water due to the attenuation of electromagnetic waves [3, 4]. Among these underwater systems, a long baseline (LBL) [5] system is worth considering since it shares similarity with a GPS in terms of configuration and working principles.

In a LBL system, a navigation subject estimates its own position using the ranges between itself and each of LBL transponders. To obtain a range, the subject receives an acoustic wave sent by a transponder. This exchange can be arranged in a two-way travel time (OWTT) or one-way travel-time (OWTT) configuration [6]. Here, a range equals to time required by the wave to travel between these two points multiplied by the wave propagation speed. This method is known as the time of flight (ToF) measurement [7]. Using several ToFs as references, the subject’s position could be then determined by applying certain technique such as time-of-arrival (ToA) [8] or time-difference-of-arrival (TDoA) [9].

An unbiased ToF measurement holds to the following ideals. First, the wave propagates in a line of sight (LoS), i.e. in a straight path. Second, the navigation subject remains still throughout the measurement. Corollary, applying either TDoA or ToA hold the same ideal. Otherwise, a ToF will result in a *pseudorange*, i.e. a range plus its bias(es). Regarding to the first one, biased ToF measurements due to non-LoS scenario may occur both in underwater acoustic [10] and electromagnetic [11] based systems.

On the other hand, a ToF differing from the second ideal would virtually affect only the acoustic based systems. This is because the propagation speed of an acoustic wave could be very low (around 1500 m/s) compared to the speed of an electromagnetic wave (\(3 \times 10^8\) m/s). Due to this characteristic, a moving navigation subject like an autonomous underwater vehicle (AUV) may have displaced while carrying out a ToF. Furthermore, ToFs may be completed in different time epochs considering the range differences among the transponders with respect to the AUV. Hence, applying either ToA or TDoA to determine position should also deal with this uncertainty.

The asynchronous ToFs in LBL navigation have been addressed in the literature. In [12, 13, 14], the AUV processes the ToFs are in a sequence with known queuing times. On the other hand, in [15] this uncertainty is a consequence of the slow acoustic wave propagation speed that occurs in a scenario where a LBL with \(10^3\,{\mathrm {m}}^2\) order of area is being used by an AUV with a considerable fast speed.

In this paper, we present LBL navigation for an AUV that deals with uncertainty due to motion of the AUV. Here, we model the state estimator while considering range bias that occurs in each ToF. Similar to [15], we also consider the clock offset as the source of range uncertainty while set a departure by addressing it now as a time-varying case. Once we obtain the systems state space representation, we apply the Kalman filter to estimate the AUV position and speed.

### 1.1 Notation

In this paper, \({\mathbf {I}}\) and \({\mathbf {0}}\) denote identity and zero matrices, respectively. Depending on the context of usages, their dimensions would adjust accordingly.

## 2 Problem statement

We consider ToF measurements involving a LBL formed by five transponders with known and fix position, \({\mathbf {r}}_1, \ldots , {\mathbf {r}}_5 \in {\mathbb {R}}^3\) and an AUV, as depicted in Fig. 1. The motion of the AUV along trajectory AB is represented by yellow-dotted green circles. To estimate its position at time *t*, namely \({\mathbf {r}}(t)\), the AUV is supposed to receive acoustic waves from all transponders at the same time, represented by the red dashed arrows. However, the acoustic waves arrive in different time epochs as results of the AUV motion, as represented by the solid black arrows.

The above scenario would be addressed with the following assumptions. First, the AUV movement is a linear motion. Second, information about senders identity and send time is encoded into each transmitted wave [16]. Third, the dynamics and uncertainties of the transponders are all identical. Fourth, the AUV position is represented by its receivers coordinate frame.

### 2.1 Vehicle kinematics

### 2.2 Pseudorange

*j*(\(j=1,\ldots ,5\)) at time sampling

*k*can be expressed as:

*c*denotes the acoustic wave speed, \(t_0(k)\) and \(t_j(k)\) denote the acoustic waves send and receive times

*k*, respectively; \(\theta (k)\) denotes transponder’s clock offset; and \(\delta _j(k)\) is a Gaussian additive noise representing pseudorange

*j*uncertainty due to the AUV movement. One may rewrite (5) as

*w*(

*k*) is additional Gauss noise for the clock offset. The dynamics of clock skew is modeled as

### 2.3 State definition

### 2.4 Pseudorange differences

*i*also denotes the index of the pseudorange (\(i=1,\ldots ,5\)) but \(i \ne j\). This technique is introduced in [8] as an approach to the ToA location. For

*L*pseudoranges, this computation will result in \(L(L-1)/2\) possible combinations.

### 2.5 State space representation

### 2.6 System observability

We are about to determine the entries of \({\mathbf {C}}(k)\) in (15) in the sense of minimum sensors, i.e. the least numbers and kinds of sensors required by the system to be observable. For this purpose, we apply the so-called graphic approach [21] by drawing an inference diagram as shown in Fig. 2. Here, every state is represented by a node. For simplicity, the index *k* is excluded from the node.

By inspecting (9)–(12) while noticing that (13) forms \(x_5(k), \ldots ,x_{14}(k)\), one would find that \({\mathbf {x}}_2(k)\), \(x_3(k)\), and \(x_4(k)\) appear on the right side of other states. Accordingly, on Fig. 2 we draw arrows toward nodes \({\mathbf {x}}_2\), \(x_3\), and \(x_4\) from other nodes. For example, we draw arrows toward node \({\mathbf {x}}_2\) from nodes \({\mathbf {x}}_1\) and \(x_5, \ldots ,x_{14}\), since appears on the right side of (9) and (13). This means that information about \({\mathbf {x}}_2(k)\) can be infer once information about \({\mathbf {x}}_1(k)\) and \(x_5(k), \ldots ,x_{14}(k)\) are available, implying that sensor deployment to measure \({\mathbf {x}}_2(k)\) is not necessary. In this sense, \({\mathbf {x}}_2(k)\) is assigned as non-sensor node. Applying the same reasoning, one would also find that \(x_3\) and \(x_4\) fall into the non-sensor node category.

### 2.7 Proposed solution

The system defined by (14)–(15) is linear, despite that some of the entries of \({\mathbf {A}}(k)\) are time-varying. Furthermore, all systems uncertainties are modeled as Gaussian. Therefore, we chose a standard Kalman filter to estimate the states.

## 3 Simulation

### 3.1 Setup and numerical values

We consider that the above scenario occurs in the middle of an extensive shallow sea. Here, five transponders with low update rate, i.e. 0.1 kHz [22], are deployed at coordinates \({\mathbf {r}}_1=\left[ \begin{array}{ccc} 0&0&0 \end{array}\right] ^T\) m, \({\mathbf {r}}_2=\left[ \begin{array}{ccc} 0&0&100 \end{array}\right] ^T\) m, \({\mathbf {r}}_3=\left[ \begin{array}{ccc} 2000&0&100 \end{array}\right] ^T\) m, \({\mathbf {r}}_4=\left[ \begin{array}{ccc} 0&2000&101 \end{array}\right] ^T\) m, and \({\mathbf {r}}_5=\left[ \begin{array}{ccc} 2000&2000&99 \end{array}\right] ^T\) m, respectively. Accordingly, the LBL update rate is set as the value of \(\tau\), i.e. 10 s.

The AUV was set to start initially at \({\mathbf {r}}(1)=\left[ \begin{array}{ccc} 340&300&5 \end{array}\right] ^T\) m and expected to finish at \({\mathbf {r}}(n)=\left[ \begin{array}{ccc} 340&300&60 \end{array}\right] ^T\) m with initial speed \({\mathbf {v}}(0=\left[ \begin{array}{ccc} 4&0&0 \end{array}\right] ^T\) m/s, while following a trajectory shown in Fig. 3. Accordingly, initial values for position and velocity estimations are set to be \(\tilde{{\mathbf {r}}}(0)=\left[ \begin{array}{ccc} 300&300&5 \end{array}\right] ^T\) m and \(\tilde{{\mathbf {v}}}(0)=\left[ \begin{array}{ccc} 4&0&0 \end{array}\right] ^T\) m/s, respectively. Meanwhile, the ToF uncertainties due to the AUV movement are depicted in Fig. 4.

*w*(

*t*)—is set to be exhibited during \(t_0(k)\), and \(\eta (k)\) are depicted in Fig. 5.

For the Kalman Filter, the *a priori* error covariance matrix is set initially to be \({\mathrm {diag}}\left( 5 \cdot {\mathbf {I}},\right.\) \(10^{-1}\cdot {\mathbf {I}},\) \(10^{-2},\) \(10^{-10},\) \(\left. 5\cdot {\mathbf {I}} \right) \in {\mathbb {R}}^{18 \times 18}\), where the entries denote the error covariance of the position, speed, clock offset, clock skew, and range differences, respectively. Subsequently, the noise covariance matrix is set to be \({\mathrm {diag}}\left( 10^{-4} \cdot {\mathbf {I}}, \right.\) \(10^{-6},\) \(10^{-4} \cdot {\mathbf {I}},\) \(10^{-5}, 10^{-12},\) \(\left. 10^{-13},10^{-6}\cdot {\mathbf {I}} \right) \in {\mathbb {R}}^{18 \times 18}\). As one may notice, smaller values are chosen for error noise covariance of the vertical position and speed entries. Finally, the measurement noise covariance matrix is set to be \({\mathrm {diag}}\left( 10^{-8} \cdot {\mathbf {I}} \right) \in {\mathbb {R}}^{10 \times 10}\).

### 3.2 Results and discussion

The simulation is run for \(n=336\), i.e 3360 s. The errors of the AUV position estimation in three dimensions are shown in Fig. 6, while the errors of the AUV velocity estimation are shown in Fig. 7, also in three dimensions. Despite exhibiting large errors at some points, in general, the estimation can follow the actual position trajectory in considerable good accuracy. Similar results are also obtained in velocity estimations. On the other hand, noticeable white noise in *z* axis is due to the fact that AUV velocity in this axis are zero in most occasions and very small in other ones.

Furthermore, compensation of the clock offset and clock skew are shown in Fig. 8. As one may notice, the peak of the clock offset contribute to more than 600 m of pseudorange. Nonetheless, the clock offset and clock skew decay over the time and eventually converge to small number.

*x*axis. Better results are obtained for

*y*and

*z*axes, where the convergence are reached after 1000 s. However, a more proper tuning on the Kalman filter seem necessary such the estimation converge faster.

## 4 Conclusion

In this paper, we have presented compensation of asynchronous ToFs in LBL navigation for an AUV. We derived model that considering the movement during the ToF as a source of uncertainties. By using graphic approach, we demonstrated the observability of the systems and minimum sensors required for the system realization. By simulation, we show that the derived model and the estimator can follow the actual position and speed of the AUV in a good accordance.

Since the update rate of the LBL is considerable very low, inclusion of a much faster sensor like inertial measurement unit (IMU) to the system becomes a sensible future work. In such configuration, the AUV would use an IMU as the main navigation aid. Noticing that an IMU tends to accumulate errors [23], the LBL would be assigned as its periodic corrector. On the other hand, how the estimator would deal with possible occasions when a ToF measurement fails would be also worth to investigate in the future.

## Notes

### Acknowledgements

Yohannes S.M. Simamora was supported by Lembaga Pengelola Dana Pendidikan (LPDP) Indonesia under contract no. PRJ-5786/LPDP.3/2016.

### Compliance with ethical standards

### Conflict of interest

On behalf of all authors, the corresponding author solemnly declares that there are no conflicts of interest

## References

- 1.Tyrrell WA (1966) Underwater acoustic positioning: principles and problems. In: First marine geodesy symposium, Columbus, OH, pp 83–92Google Scholar
- 2.Spiess FN (1966) Underwater acoustic positioning: applications. In: First marine geodesy symposium, Columbus, OH, pp 93–101Google Scholar
- 3.Al-Shamma’a AI, Shaw A, Saman S (2004) Propagation of electromagnetic waves at MHz frequencies through seawater. IEEE Trans Antennas Propag 52(11):2843–2849CrossRefGoogle Scholar
- 4.Hovem JM (2007) Underwater acoustics: propagation, devices and systems. J Electroceram 19(4):339–347CrossRefGoogle Scholar
- 5.Van Calcar H (1969) Acoustic position reference methods for offshore drilling operations. In: Offshore technology conference, Houston, TXGoogle Scholar
- 6.Almeida R, Melo J, Cruz N (2016) Characterization of measurement errors in a LBL positioning system. In: OCEANS 2016, Shanghai, China, pp 1–6Google Scholar
- 7.Bensky A (2007) Wireless positioning technologies and applications. Artech House, NorwoodGoogle Scholar
- 8.Caffery JJ (2000) A new approach to the geometry of TOA location. In: Proceedings of IEEE VTS Fall VTC2000 52nd vehicular technology conference, Boston, MA, vol 4, pp 1943–1949Google Scholar
- 9.Potluri S (2002) Hyperbolic position location estimator with tdoas from four stations. Master’s thesis, New Jersey Institute of TechnologyGoogle Scholar
- 10.Ramezani H, Jamali-Rad H, Leus G (2013) Target localization and tracking for an isogradient sound speed profile. IEEE Trans Signal Proc 61(6):1434–1446MathSciNetCrossRefGoogle Scholar
- 11.Breßler P, Reisdorf P, Obst M, Wanielik G (2016) GNSS positioning in non-line-of-sight context—a survey. In: 2016 IEEE 19th ITSC, Rio de Janeiro, pp 1147–1154Google Scholar
- 12.Miller PA, Farrell JA, Zhao Y, Djapic V (2010) Autonomous underwater vehicle navigation. IEEE J Ocean Eng 35(3):663–678CrossRefGoogle Scholar
- 13.Gode T (2015) Long baseline ranging acoustic positioning system. Master’s thesis, Virginia Polytechnic Institute and State UniversityGoogle Scholar
- 14.Chen Y, Zheng D, Miller PA, Farrell JA (2016) Underwater inertial navigation with long baseline transceivers: a near-real-time approach. IEEE Trans Control Syst Technol 24(1):240–251CrossRefGoogle Scholar
- 15.Yi J, Mirza D, Kastner R, Schurgers C, Roberts P, Jaffe J (2015) ToA-TS: time of arrival based joint time synchronization and tracking for mobile underwater systems. Ad Hoc Netw 34:211–223CrossRefGoogle Scholar
- 16.Webster SE, Eustice RM, Singh H, Whitcomb LL (2012) Advances in single-beacon one-way-travel-time acoustic navigation for underwater vehicles. Int J Robot Res 31(8):935–950CrossRefGoogle Scholar
- 17.Simamora YSM, Tjokronegoro HA, Leksono E (2018) Modeling of long baseline navigation in the presence of time-varying clock-offsett. In: ICETASIA, Surakarta, IndonesiaGoogle Scholar
- 18.Langley RB (1999) Dilution of precision. GPS World 10:52–59Google Scholar
- 19.Kim H-Y (2014) Modeling and tracking time-varying clock drifts in wireless networks. PhD thesis, Georgia Institute of TechnologyGoogle Scholar
- 20.Batista P (2015) Long baseline navigation with clock offset estimation and discrete-time measurements. Control Eng Pract 35:43–53CrossRefGoogle Scholar
- 21.Liu Y-Y, Slotine J-J, Barabási A-L (2013) Observability of complex systems. Proc Natl Acad Sci 110(7):2460–2465MathSciNetCrossRefGoogle Scholar
- 22.Kinsey JC, Yang Q, Howland JC (2014) Nonlinear dynamic model-based state estimators for underwater navigation of remotely operated vehicles. IEEE Trans Control Syst Technol 22(5):1845–1854CrossRefGoogle Scholar
- 23.Noureldin A, Karamat TB, Georgy J (2013) Fundamentals of inertial navigation, satellite-based positioning and their integration. Springer, BerlinCrossRefGoogle Scholar