Professional Documents
Culture Documents
"RCC11 A Finer Topological Representation For The Alignment of Regions in Sketch Maps" - Jan, S. Schwering, A. Schultz, C. Chipofya, M. (2015)
"RCC11 A Finer Topological Representation For The Alignment of Regions in Sketch Maps" - Jan, S. Schwering, A. Schultz, C. Chipofya, M. (2015)
"RCC11 A Finer Topological Representation For The Alignment of Regions in Sketch Maps" - Jan, S. Schwering, A. Schultz, C. Chipofya, M. (2015)
Abstract Introduction
During the last two decades, dozens of qualitative
The main goal of qualitative representation and reasoning
representations have been proposed. These representations
is to represent our everyday commonsense knowledge
are motivated by a wide variety of applications of spatial
about the physical world. It provides mechanisms which
data processing such as Geographical Information System
characterize essential properties of elementary objects (e.g.
(GIS), robotic navigation, and high level vision. For
points, lines, and regions) and spatial configurations
topological reasoning, the Region Connected Calculus
between them (e.g. adjacent, on the left of, and included
(RCC) is perhaps the best-known formalism. The two
in).
algebras, RCC5 and RCC8 distinguish five and eight
In the area of Qualitative Spatial Reasoning (QSR),
different topological relations. The different levels of
dozens of spatial representations have been proposed
granularity provide flexibility in the selection of
focusing on different aspects of the space such as
representations suitable for different applications. In this
representations for topology (Randell, Cui, and Cohn 1992;
paper, we propose RCC11, finer versions of the RCC in
Cohn et al. 1997), directions (Frank 1996; Renz and Mitra
which topological relations are refined using the geometric
2004), relative position of points (Moratz, Dylla, and
point-set approach. If one region is tangential proper part
Frommberger 2005; Renz and Mitra 2004; Moratz, Renz,
of another or if two regions externally connect then their
and Wolter 2000) and others, each of which introduces a
topological relation is further distinguished by the
finite number of basic spatial relations. The qualitative
dimension of the intersection of their boundaries: line or
representations and methods for reasoning are motivated
point contact. This is an important spatial distinction for
by a wide variety of application areas such as Geographical
users of GIS to query and retrieve information from
Information System (GIS), robotic navigation, engineering
databases. The qualitative qualifier is used to compute the
design, and commonsense reasoning about the physical
RCC11 topological relation between closed regions. The
world.
composition table for base relations is computed using the
In the context of GIS, the spatial relations between the
declarative spatial reasoning system CLP(QS). The
geographic objects play a central role in query
proposed representation is evaluated within the application
specification, query processing and for retrieval tasks. For
of sketch map alignment: We compute regions (city-
modelling topological relations in GIS and spatial
blocks) in sketch and geo-referenced maps. First, RCC11
databases the three models: the 9-intersection (9IM)
relations between city-blocks are extracted in the form of
(Egenhofer, Franzosa, and Ranzosas 1991), RCC-family
qualitative constraint networks. Afterwards, the evaluation
(Randell, Cui, and Cohn 1992; Cohn 1997), and Calculus
of RCC11 is done by matching the qualitative constraint
Based Method (CMB) (Clementini, Felice, and Oosterom
networks from sketch and geo-referenced maps.
1993) play an important role both in terms of theoretical
Keywords: qualitative representations, sketch maps, developments and practical applications. In many geo-
qualitative alignments, geo-referenced maps, region spatial applications (Egenhofer 1996; Forbus et al. 2011;
connection calculus Volker and Michael 1997; Nedas and Egenhofer 2008),
these representations are used to make qualitative
distinctions and process spatial information on a qualitative
level.
Copyright © 2015, Association for the Advancement of Artificial In our previous work (Schwering et al. 2014; Chipofya,
Intelligence (www.aaai.org). All rights reserved. Wang, and Schwering 2011), we propose a theoretical
framework to use freehand sketches as a visual interaction the matching of spatial objects, where the composition
mechanism to process spatial information qualitatively. In table of the representation helps to prune the search space
freehand sketches, street segments are linear features. They and perform the consistency check during constraint–based
are connected to other street segments at junctions. reasoning. The overall evaluation shows that the finer
Landmarks represent all other geographical objects such as topological relations using RCC11 are suitable for the
water bodies, landmarks, and parks. City-blocks are the alignment of spatial objects in both sketch and geo-
smallest regions. They are delimited by a lineal referenced maps.
representation of connected street segments. As typically The remainder of this paper is structured as follows: in
only qualitative relations between these spatial objects are the following section, we introduce the background and
persevered in sketch maps, processing spatial information related work on sketch maps representation and RCC
on a qualitative level has been suggested (Egenhofer 1996; theory. In section 2, we describe the RCC11 relations, and
Chipofya, Wang, and Schwering 2011). In (Jan et al. give an illustration of topological relations between closed
2014a; Jan et al. 2014b; Jan et al. 2014c; Schwering et al. regions. In section 3, we discuss the evaluation of the
2014), we proposed a set of plausible qualitative proposed representation against the RCC8 representation
representations to formalize the spatial aspects preserved in through qualitative matching. Section 4 concludes the
sketch maps. paper with an outlook on future work.
This study extends our previous work on qualitative
representations for the alignment of spatial objects. In this
paper, we propose a finer representation, called RCC111, to Background and Related Work
formalize topological relations between city-blocks. The
RCC11 distinguishes 11 topological relations between Sketch Maps Representation
closed regions in the plane ℝ2 based on the dimension of People capture the spatial configurations of a scene in a
the intersection of boundaries which we distinguish into hand-drawn map, called a sketch map. As this information
line or point contact. The RCC8 captures a very general is based on observations rather than measurements, it is
notion of connectivity, which may be useful for various distorted, schematized, incomplete and generalized
purposes. However, the representation appears too weak to (Tversky 1992). People draw only a few significant objects
formalize the important topological distinctions of such as landmarks and the street network composed of
connectivity such as regions being connected by lines or street segments, junctions, and city-blocks.
points. Using the proposed representation, we capture the During the last two decades, several approaches attempt
important topological distinctions that are preserved in to process spatial information from sketch maps at a
sketch maps such as city-blocks being externally connected qualitative level. Egenhofer (Egenhofer 1996) proposed
by street segments (i.e. line segment contact) or being “Spatial-Query-by-Sketch”, an approach to query spatial
connected diagonally at junctions (i.e. point contact), databases using a sketch-based interface. The approach
which are important distinctions for qualitative alignment. uses topological relations (9-intersection model) and
The composition table specifies the relations obtained by cardinal relations to formalize the sketched scene. Volker
composing the 11 base relations. It provides the basis for et al. (Volker and Michael 1997) propose the visual query
composition-based reasoning, extensively used in relation system—VISCO. It integrates geometric and topological
algebraic qualitative spatial reasoning. querying with deductive spatial reasoning. Forbus et al.
The proposed representation is evaluated by aligning (Forbus et al. 2011) develop a sketch understanding
city-blocks from sketch maps with the corresponding city- system—nuSketch. It uses both qualitative topological
blocks in geo-referenced maps. The qualitative matching of reasoning and quantitative information to construct spatial
spatial objects requires structures called Qualitative configurations between depicted objects. Nedas et al.
Constraint Networks (QCNs), which represent pairs of (Nedas and Egenhofer 2008) propose a similarity measure
spatial objects and relations among objects for a particular to compare two spatial scenes by identifying similarities
aspect of space. We use the qualitative qualifier (Jan and between (i) objects in scenes, (ii) relations among objects,
Chipofya 2011) to compute QCNs using RCC11 from the and (iii) the ratio of matched objects.
geometric representations of both sketch and geo- In empirical studies (Wang, Muelligann, and Schwering
referenced maps. Afterwards, the QCNs are used as inputs 2011; Schwering et al. 2014; Wang, Mülligann, and
in qualitative matching algorithms (Wallgrün, Wolter, and Schwering 2010), we determined a set of qualitative
Richter 2010; Chipofya, Schwering, and Binor 2013). Both relations between depicted objects that are not affected by
matching approaches use constraint-based techniques for schematization and distortions, and are usually represented
correctly in sketch maps. The topological relation between
1
The topological relations presented in this paper are different from the city-blocks is one of these qualitative relations that is of
11 relations introduced in (Düntsch 1999).
particular interest for this paper; the complete list of An Overview of RCC Theory
qualitative relations can be found in (Schwering et al. The formal definition of geometric objects and relations
2014). In sketch maps, city-blocks have significant are based on the point-set approach, where features are sets
topological distinctions such as disconnect, externally and points are elements of the sets (Clementini, Felice, and
connected by line (sharing common street segments), and Oosterom 1993). In this way, all the geometric features are
externally connected by point (sharing common junctions). closed sets where each feature contains all its accumulation
However, city-blocks never overlap each other as they are points. The spatial features used in GIS are simple points,
delimited by connected street segments. lines, and regions in the plane ℝ2 .
People do not always sketch complete city-blocks, in The focus of qualitative representation and reasoning on
particular at the edge of the sketch medium. We define these features started with the evaluation and
city-blocks as areas either bounded by the street segments, implementation of formal axioms for space and time
or bounded by street segments and the sketch-boundary (Clarke 1981) and expressed in the many sorted logic
(Jan et al. 2014a). As shown in the Figure 1b, all LLAMA (Cohn 1987). Based on “calculus of individuals
incomplete street segments are extended towards the based on connection” (Clarke 1981), the RCC theory is
sketch-boundary. We extracted all depicted spatial objects developed throughout a series of research work (Randell,
from sketch maps using the segmentation procedures Cui, and Cohn 1992; Cohn 1995; Cohn 1997; Cohn et al.
proposed in (Broelemann 2011; Broelemann, Jiang, and 1997; Bennett, Isli, and Cohn 1998). The most distinctive
Schwering 2011). feature between Clarke´s theory and RCC theory is that
RCC theory considers extended regions rather than points
as fundamental.
The basic part of the RCC theory assume a primitive
dyadic relation called 𝒞(𝓍, 𝓎), which represents the
connectivity of two regions 𝓍 and 𝓎 and the relation
𝒞(𝓍, 𝓎) is reflexive and symmetric. Using the 𝒞(𝓍, 𝓎),
further dyadic relations are defined such as: DC
(disconnect), P (part of), PP (proper part), PO (partially
overlap), EC (externally connected), TPP (tangential
Figure 1: (a) Sketch map, (b) the depicted street segments, proper part), NTPP (non-tangential proper part), EQ
landmarks (a1, a2,... a8), and city-blocks (cb1, cb2, cb3,… cb7) (equal), O (overlay), DR (discrete) and the converse
bounded by the street segments and medium-boundary in sketch relations of the P, PP, TPP, and NTPP.
map. A region is a set 𝐴 in the plane ℝ2 if it is non-empty and
regular closed, i.e. 𝐴 = 𝐴° ≠ ∅, where 𝑥° and 𝑥 represent
In order to formalize the topology of city-blocks, we
the interior and respectively, the closure of a set 𝑥. If we
(Jan et al. 2014a) analyze and evaluate different
have two regions 𝐴 and 𝐵, the RCC8 base relations will be
representations of RCC (Randell, Cui, and Cohn 1992;
as follows:
Cohn 1997) and string-based topological relations (Li and
Liu 2010). As a preliminary result, we find string-based (A, B) ∈ 𝐃𝐂 if A ∩ B = ∅
topological relations as a reliable model to capture the (A, B) ∈ 𝐄𝐂 if A° ∩ B° = ∅ but A ∩ B ≠ ∅
topological relations between depicted city-blocks. The (A, B) ∈ 𝐏𝐎 if A° ∩ B° ≠ ∅ and A ⊈ B and B ⊈ A
representation distinguishes the external connectivity of (A, B) ∈ 𝐓𝐏𝐏 if A ⊆ B but A ⊈ B°
city-blocks by street segments and junctions, which is an (A, B) ∈ 𝐍𝐓𝐏𝐏 if A ⊆ B°
important distinction for qualitative alignment. Using (A, B) ∈ 𝐄𝐐 if A = B
RCC8 (Randell, Cui, and Cohn 1992; Cohn 1997), we lose
The above basic relations together with the converses of
the distinction as the single relation “EC” of the RCC8
TPP and NTPP are jointly exhaustive and pairwise
represents both scenarios.
disjoint (JEPD).
However, string-based topological relations (Li and Liu
2010) are restricted to convex regions and each topological
scenario can be represented using multiple sets of circular Other Topological Representations
strings. The qualitative alignment of spatial objects In (Egenhofer and Herring 1991; Egenhofer, Franzosa, and
involves spatial reasoning. Therefore, we need a set of Ranzosas 1991), Egenhofer et al. propose a formal
tractable relations which are mutually exclusive. In approach to define the binary topological relations between
freehand sketches, city-blocks are mixed closed regions objects based on point-set theory. A drawback of the
(concave and convex). Therefore, we need a representation approach is that it distinguishes only between empty and
which supports both concave and convex regions. non-empty intersections of boundaries and the interior of
the geometries and the method also results in too many (A, B) ∈ 𝐃𝐂 if ∂A ∩ ∂B = ∅
different topological relations for end users. (A, B) ∈ 𝐄𝐂𝐩 if A° ∩ B° = ∅ but 𝑑𝑖𝑚 (∂A ∩ ∂B) = 0
Clementini et al. (Clementini, Felice, and Oosterom (A, B) ∈ 𝐄𝐂𝐥 if A° ∩ B° = ∅ but 𝑑𝑖𝑚 (∂A ∩ ∂B) = 1
1993) propose an approach called Calculus Based Method (A, B) ∈ 𝐏𝐎 if 𝑑𝑖𝑚(A° ∩ B°) = 2 ⋀ dim (∂A ∩ ∂B) =
(CBM) for representing a small set of topological relations 0 ⋀ 𝑑𝑖𝑚(A° ∩ ∂B) = 1 ⋀ 𝑑𝑖𝑚(∂A ∩ B°) = 1
based on the dimension of the intersecting geometries. The (A, B) ∈ 𝐓𝐏𝐏𝐩 if 𝑑𝑖𝑚 (A° ∩ B°) = 2 ⋀ 𝑑𝑖𝑚 (∂A ∩
resulting relations are grouped together into a few general ∂B) = 0 ⋀ 𝑑𝑖𝑚(A° ∩ ∂B) = ∅ ⋀ 𝑑𝑖𝑚(∂A ∩ B°) = 1
topological relations such as touch, in, cross, overlap, and
(A, B) ∈ 𝐓𝐏𝐏𝐥 if 𝑑𝑖𝑚(𝐴° ∩ 𝐵°) = 2 ⋀ 𝑑𝑖𝑚 (∂A ∩
disjoint.
∂B) = 1 ⋀ 𝑑𝑖𝑚(A° ∩ ∂B) = ∅ ⋀ 𝑑𝑖𝑚(∂A ∩ B°) = 1
(A, B) ∈ 𝐍𝐓𝐏𝐏 if 𝑑𝑖𝑚(𝐴° ∩ 𝐵°) = 2 ⋀ 𝑑𝑖𝑚(∂A ∩
RCC11 Base Relations ∂B) = ∅ ⋀ 𝑑𝑖𝑚(𝐴° ∩ 𝜕𝐵) = ∅ ⋀ 𝑑𝑖𝑚(𝜕𝐴 ∩ 𝐵°) = 1
(A, B) ∈ 𝐄𝐐 if 𝑑𝑖𝑚(𝐴° = 𝐵°) = 2 ⋀ dim(𝜕𝐴 = 𝜕𝐵) = 1
In the context of Geographic Information Systems (GISs),
the spatial relationships existing between geographic The above basic relations together with the converses of
objects play a central role both at the spatial queries 𝐓𝐏𝐏𝐩 , 𝐓𝐏𝐏𝐥 , and 𝐍𝐓𝐏𝐏 are the set of finer topological
definition and processes. The two algebras relations. Using the above definitions, the RCC11 has the
(RCC5 and RCC8) of RCC make simple and general following mutually exclusive relations (see Figure 2):
classifications, however, these classifications may not be
precise enough when fine grained information is required, Name of the RCC11 relations Symbol
for example, when information about the dimensionality of Disconnect DC
intersections between regions is required. Equal EQ
In this paper, we propose RCC11, a new representation Partially Overlay PO
combining the ideas of finer topological relations in Externally Connected by point ECp
(Clementini, Felice, and Oosterom 1993), the concept of Externally Connected by line ECl
strong connection and congruence in (Borgo, Guarino, and Tangential Proper Part by point TPPp
Masolo 1996), and string-based topological relations Tangential Proper Part by line TPPl
between convex regions (Li and Liu 2010). For each pair Non-Tangential Proper Part NTPP
of regions (A, B) in the plane, RCC8 classifies topological Tangential Proper Part by point inverse TPPpi
relations based on empty and non-empty intersections Tangential Proper Part by line inverse TPPli
without distinguishing certain cases of intersections of the Non-Tangential Proper Part inverse NTPPi
boundaries of two regions leading to refinements of the
Table 1: The 11 topological relations in RCC11.
RCC8 relations EC, TPP, and TPPi. In our approach, we
take into account the dimension of intersections
Composition Table
(Clementini, Felice, and Oosterom 1993) and line-point
contact instead of only considering empty or non-empty Originating from Allen´s interval Algebra (IA),
intersections. In 2D-dimensional space, the intersection set composition-based reasoning (Guesgen 1989) has been
(I) can be either empty (∅), point (0D), line (1D), or region widely acknowledged as the most popular reasoning
(2D). The dimension of the intersection cannot be higher technique in the area of QSR. Given a fixed vocabulary of
relations, the composition table enables us to infer implicit
than the lowest dimension of the two intersecting
qualitative knowledge. A difficult problem in QSR is the
operands: 𝑑𝑖𝑚 (𝜕𝐴) = 1 and 𝑑𝑖𝑚 (𝐴°) = 2. For the
computation of a composition table, the verification that all
topological relations between two closed region A and B
the closure properties are met, and the determination of the
(with their interiors°, and boundaries 𝜕), we distinguish the computational complexity.
following intersections: Deriving the composition table manually is a
I = (∂A, ∂B) ∶ ∅, 0D, 1D (3 cases) challenging and error-prone process, particularly if the
representation contains many base relations. Thus, in order
I = (∂A, B°) ∶ ∅, 0D, 1D (3 cases)
to automatically generate the composition table, we use the
I = (A°, ∂B) ∶ ∅, 0D, 1D (3 cases) declarative spatial reasoning system CLP(QS) (Schultz and
I = (A°, B°) ∶ ∅, 2D (2 cases) Mehul 2014; Schultz and Mehul 2012; Bhatt, Lee, and
Schultz 2011) refer to Appendix A for further details. The
Using the dimension of the intersections between composed relations in the composition table (see Appendix
regions, we have 11 possible topological relations. The “p” B) are refined relations ℛ(𝑥, 𝑦) that hold between any two
and “l” in relations represent the dimension (dim) of variables 𝑥 and 𝑦 occurring in the network via the
intersection by point or line, respectively. following operation.
ℛ (𝑥,𝑦) ← ℛ(𝑥,𝑦) ∩ (ℛ(𝑥,𝑧) ° ℛ(𝑧,𝑦) )
The composition table enables topological reasoning at
the conceptual level, rather than having to calculate all
relations from the geometrical representation of the spatial
objects in the ℝ2 . We use the composition table as a
computational model to assess the consistency of the
topological relations between city-blocks during the
alignment.
ecl ecp
ntpp tpp
ecl p tppl ecp
Figure 4: Tubular model for RCC11 weak composition.
ntpp tppp
For example, consider proving the theorem: there exists tpp
contact; the topological relations between the bases
l
circles a, b, and c such that DCab ⋀ TPPibc ⋀ ECac. This distinguishes the standard RCC8 contact and containment
corresponds to the constraint system: there exists reals xa, relations.
ya, ra, xb, yb, rb, xc, yc, rc such that: The topological relations between tubes tab, t’a’b’ are as
follows:
(xa - xb)2 + (ya - yb)2 > (ra + rb)2 (DCab)
(xc - xb)2 + (yc - yb)2 = (rc - rb)2 (TPPibc) DCt t’ ≡def DCa a’
rc < rb ntpp ECp t t’
tpp≡pdef ECa a’ ⋀tpp
DCb b’l
(xa - xc)2 + (ya - yc)2 = (ra + rc)2 (ECac) ECl t t’ ≡def ECa a’ ⋀ ECb b’
POt t’ ≡def POa a’
CLP(QS) determines that this system of constraints is TPPp t t’ ≡def TPPa a’ ⋀ NTPPb b’
unsatisfiable, and therefore the spatial relations are
inconsistent. Thus, EC must not appear in the RCC8
TPPl t t’ ≡def TPPa a’ ⋀ TPPb b’
composition table cell for row DC and column TPPi. NTPPt t’ ≡def NTPPa a’ ⋀ NTPPb b’
As circles are a representation for RCC8, they can be TPPip t t’ ≡def TPPp t’t
used to determine the ways that three objects can have TPPilt t’ ≡def TPPl t’t
contact without specifying the nature of that contact. To NTPPit t’ ≡def NTPPt’t
distinguish between line and point contact we build on the EQt t’ ≡def EQa a’ ⋀ EQb b’
circular model by proposing our 3D tubular model for
RCC11 weak composition. Let tube tab be a pair of circles It is straightforward to show that these topological
a, b such that b is part of a (i.e. Pba): relations between tubes satisfy our RCC11 base relation
definitions (in particular, with respect to the dimension of
Pba ≡def Δ(ca,cb) ≤ (ra - rb)2 ⋀ (rb ≤ ra) contact). CLP(QS) is used to prove (or disprove) all
113 = 1331 composition table theorems.
Topological relations are defined by comparing the base
circles and top circles between tubes as illustrated in
Figure 4. The interpretation is that the tubes are hollow,
they do not have base nor lid covers, they are sitting on the
same plane, they all have the same height, and all tops lie
in a plane parallel to the base plane. The tubes can be
negatively tapered, i.e. they taper inwards, and the axis
does not need to be at right angles to the plane (formally,
they are axis-aligned open circular cylinders that can
inwardly taper to become cones). The contact between the
surfaces of tubes distinguishes between line and point
Appendix B: The RCC11 Composition Table
A (R1) B
DC ECp ECl EQ PO TPPp TPPl TPPpi TPPIi NTPP NTPPi
B (R2 ) C
DC, ECp, ECl DC, ECp, ECl, DC, ECp, ECl, DC, ECp, ECl, DC, ECp, ECl, DC, ECp, ECl,
DC .-?-. PO, TPPp, PO, TPPp, DC PO, TPPp, PO, TPPp, PO, TPPp, DC DC PO, TPPp, DC
TPPl, NTPP TPPl, NTPP TPPl, NTPP TPPl, NTPP TPPl, NTPP TPPl, NTPP
DC, ECp, ECl,
DC, ECp, ECl, DC, ECp, ECl, ECp, ECl, PO, ECp, ECl, PO,
EQ, PO, TPPp, DC, ECp, ECl, PO, TPPp,
ECp PO, TPPpi, ECp PO, TPPp, TPPp, TPPl, TPPp, TPPl, DC, ECp DC, ECp DC
TPPl, TPPpi, PO, TPPp, TPPl TPPl, NTPP
TPPli, NTPPi TPPl, NTPP NTPP NTPP
TPPli
DC, ECp, ECl, DC, ECp, ECl, DC, ECp, ECl, DC, ECp, ECl,
PO, TPPp, Ecl, PO, TPPp, PO, TPPp,
ECl PO, TPPpi, PO, TPPpi, EQ, PO, TPPl, ECl PO, TPPp, DC, ECp DC, ECp, ECl DC
TPPl, NTPP TPPl, NTPP TPPl, NTPP
TPPli, NTPPi TPPli TPPli TPPl, NTPP
DC, ECp, ECl, DC, ECp, ECl, DC, ECp, ECl, DC, ECp, ECl, DC, ECp, ECl, DC, ECp, ECl,
PO, TPPp, PO, TPPp, PO, TPPp,
PO PO, TPPpi, PO, TPPpi, PO, TPPpi, PO .-?-. PO, TPPpi, PO, TPPpi, PO, TPPpi,
TPPl, NTPP TPPl, NTPP TPPl, NTPP
TPPli, NTPPi TPPli, NTPPi TPPli, NTPPi TPPli, NTPPi TPPli, NTPPi TPPli, NTPPi
DC, ECp, ECl,
DC, ECp, ECl, DC, ECp, ECl,
EQ, PO, TPPp, DC, ECp, ECl,
TPPp DC DC, ECp DC, ECp TPPp PO, TPPp, TPPp, NTPP TPPp, NTPP NTPP PO, TPPpi,
TPPl, TPPpi, PO, TPPp, TPPl
TPPl, NTPP TPPli, NTPPi
TPPli
DC, ECp, ECl, DC, ECp, ECl, DC, ECp, ECl, DC, ECp, ECl,
TPPp, TPPl,
TPPl DC DC, ECp DC, ECp, ECl TPPl PO, TPPp, TPPp, NTPP PO, TPPpi, EQ, PO, TPPl, NTPP PO, TPPpi,
NTPP
TPPl, NTPP TPPli TPPli TPPli, NTPPi